Tempris Pivotal Roles · Accounting
Sector report

Pivotal roles in accounting

Twelve roles that carry the accounting function, read task by task against what AI can take on today. An evidence baseline for where the work changes first.

Prepared by Tempris
Roles analysed 12
Mean automation potential 38%
Method Job-description evidence, task-level analysis
Why this report

Accounting is where the AI question gets concrete

Accounting work is structured and heavily rules-bound. That makes it one of the first functions where the AI conversation can move from speculation to evidence.

The questions inside a finance team are practical. Which roles change first, and by how much. Which tasks are open to AI support, and which only look that way. Where to reinvest the time that AI frees up. This report answers those questions for twelve roles that hold the core of the accounting function, from transaction processing through controllership, audit, tax, and treasury.

The report is a baseline, not a forecast. It is built from how these roles are described in practice and broken into the tasks people do week to week.

How to read this

From job description to task to capability

Every figure traces back to observed work.

Each role was taken from job-description evidence and broken into its tasks, and each task into its actions. Every action was assessed against four AI capabilities, scoring how far each could carry that piece of work and where a person still leads. Rolled up, that produces an automation potential for each task and each role: the share of measured task time open to AI support, with a person remaining the lead throughout.

Automation potential measures opportunity, not headcount. A role at 45% is not a role half removed. It is a role where close to half the task time could shift toward judgement and advisory work, if the team chooses to.

The cohort

Twelve roles, clustered tightly around 38%

Across the accounting cohort, mean automation potential is 38.1% and the median 36.8%. What stands out is how narrow the spread is.

2
30–35%
7
35–40%
2
40–45%
1
45–50%

Every one of the twelve roles lands between 33% and 50%. No role is untouched by AI, and none is close to fully exposed. Accounting does not split into safe and at-risk roles. The whole function shifts together, by a meaningful but bounded degree. The response that follows is a function-wide redesign of how the work is done.

The work, clustered

Six themes that cut across the function

The cohort's 103 tasks group into six themes of related work, each spanning several roles. Ranked here by automation potential, weighted by how much role time each theme carries.

01

Accounts Payable Management

45% automation potential

This cluster encompasses tasks related to overseeing, managing, and executing accounts payable operations and related financial processes.

17 tasks · spans 4 of 12 roles · peak task Handle accounts payable (58%)
Accounts Payable Manager7Bookkeeper4Treasury Accountant4Financial Controller2
02

Financial Reporting and Analysis

40% automation potential

This cluster encompasses tasks related to the preparation, analysis, and management of financial reports and statements to ensure accuracy and compliance.

21 tasks · spans 7 of 12 roles · peak task Ensure accurate financial reporting (55%)
FP&A Manager7Financial Controller7Treasury Accountant2Statutory Reporting Manager2Bookkeeper1Management Accountant1Senior Auditor1
03

Tax Compliance and Management

38% automation potential

This cluster encompasses tasks related to the preparation, filing, and management of tax-related activities and compliance with statutory requirements.

12 tasks · spans 6 of 12 roles · peak task Manage client portfolio (49%)
Statutory Reporting Manager4Tax Manager4Accounts Payable Manager1Bookkeeper1Management Accountant1Financial Controller1
04

Audit Management

36% automation potential

This cluster encompasses tasks related to planning, executing, and reporting on audits, ensuring compliance, and managing audit-related communications and issues.

20 tasks · spans 5 of 12 roles · peak task Assess and validate internal controls (43%)
Audit Manager9Internal Auditor6Senior Auditor3Treasury Accountant1Tax Manager1
05

Forensic Accounting and Investigations

34% automation potential

This cluster encompasses tasks related to the analysis, investigation, and reporting of financial data to uncover fraud and improve financial systems.

12 tasks · spans 4 of 12 roles · peak task Assess financial evidence (42%)
Forensic Accountant8Senior Auditor2Treasury Accountant1Management Accountant1
06

Team and Project Management

33% automation potential

This cluster encompasses tasks related to managing teams, client relationships, and strategic projects to ensure effective operations and service delivery.

21 tasks · spans 9 of 12 roles · peak task Manage stakeholder relationships (48%)
Management Accountant5Statutory Reporting Manager3Internal Auditor3Tax Manager3Bookkeeper2FP&A Manager2Accounts Payable Manager1Treasury Accountant1Senior Auditor1

Themes, not roles, are the better unit for planning. Accounts payable and financial reporting work carry the highest automation potential and recur across four to seven roles each, so an investment there pays back in several places at once. Team and project management sits lowest and spans the most roles, the connective work that stays with people. Each theme is a candidate for a single, shared redesign rather than a dozen separate role-level fixes.

Role exposure, ranked

Where the work changes first

Roles ordered by automation potential. Processing-heavy roles lead the ranking; judgement- and relationship-heavy roles sit at the foot.

Bookkeeper
49%
Financial Controller
41%
FP&A Manager
40%
Accounts Payable Manager
40%
Statutory Reporting Manager
39%
Internal Auditor
37%
Management Accountant
36%
Audit Manager
36%
Treasury Accountant
36%
Senior Auditor
36%
Forensic Accountant
35%
Tax Manager
34%

The pattern follows the nature of the work. Bookkeeping, payables, and statutory reporting are structured and repeatable, so they sit at the top. Tax, forensic, and audit work depends on client judgement and investigation, so it sits at the foot. What separates the two ends is the composition of the task profile, not the seniority of the role.

The capability mix

What is doing the work

Automation potential broken into the four AI capabilities that produce it, summed across the cohort.

Tool Use & Integration 12.2pts

Moving and reconciling data between ledgers, banking portals, and treasury or ERP systems.

Knowledge & Reasoning 11.9pts

Applying standards, interpreting figures, and working out the cause of a discrepancy or variance.

Generation & Transformation 8.1pts

Drafting statements, filings, memos, and reconciliation narratives from structured inputs.

Perception & Ingestion 5.8pts

Reading invoices, statements, and source documents, and extracting the fields that matter.

Accounting's exposure is led by tool use and reasoning. The work AI carries best here is moving and reconciling data, and applying standards to it. Content generation contributes less. That shapes how teams should adopt: the early value sits in connected, data-handling workflows.

What stays with people

The skills that hold the function together

The skills that appear across the most roles in the cohort.

Stakeholder communication10 roles
Team leadership6 roles
Client relationship management3 roles
Process improvement3 roles

Stakeholder communication appears in ten of the twelve roles. Team leadership follows, with client relationship management and process improvement close behind. The connective tissue of the accounting function is relational and judgemental. It is the part of the work the automation figures above touch least, and the part worth protecting and growing as task time is freed up.

Roles up close

Three roles across the range

The cohort's highest, mid, and lowest-exposure roles, each broken into its tasks with the share of role time and the automation potential of each.

Highest exposure in the cohort

Bookkeeper

49% task time
open to AI

Transaction-processing work fills most of this role: payables, receivables, payroll, and reconciliations. Almost all of it is structured and rules-bound, and close to half the role's measured task time is open to AI support.

Handle accounts payable
58% automation 15% of role time
Process payroll
55% automation 15% of role time
Reconcile general ledger accounts
52% automation 20% of role time
Identify and resolve discrepancies
49% automation 10% of role time
Manage accounts receivable
46% automation 15% of role time
Maintain client relationships
43% automation 5% of role time
Communicate with government agencies
41% automation 10% of role time
Assist with budgeting and planning
41% automation 10% of role time
Tool Use & Integration
22.2
Knowledge & Reasoning
11.7
Generation & Transformation
8.9
Perception & Ingestion
6.6
Mid-cohort, oversight plus operations

Financial Controller

41% task time
open to AI

The controller pairs a heavy operational core (close, ledger, day-to-day accounting) with judgement-bound oversight of compliance and controls. The operational half carries most of the role's exposure. The oversight half holds it down.

Direct day-to-day accounting operations
54% automation 14% of role time
Oversee financial statement preparation
45% automation 9% of role time
Ensure tax compliance
43% automation 9% of role time
Improve accounting processes and controls
42% automation 9% of role time
Manage chart of accounts and general ledger
39% automation 9% of role time
Perform variance and trend analyses
38% automation 9% of role time
Implement automated financial reporting systems
37% automation 9% of role time
Lead financial close processes
35% automation 14% of role time
Develop and monitor cash flow forecasts
35% automation 9% of role time
Ensure compliance with ASPE standards
33% automation 9% of role time
Tool Use & Integration
16.7
Knowledge & Reasoning
9.6
Generation & Transformation
9.4
Perception & Ingestion
4.9
Lowest exposure in the cohort

Tax Manager

34% task time
open to AI

The Tax Manager's compliance efforts sit well within reach of AI support. What holds the figure down is the share of the week spent on client judgement, planning strategy, and developing staff.

Manage client relationships
44% automation 15% of role time
Ensure compliance with standards
36% automation 5% of role time
Develop tax planning strategies
36% automation 15% of role time
Prepare and review tax returns
34% automation 20% of role time
Conduct tax research
32% automation 15% of role time
Collaborate on strategic tax initiatives
32% automation 15% of role time
Oversee engagement workflows
28% automation 5% of role time
Mentor and develop staff
18% automation 10% of role time
Knowledge & Reasoning
13.5
Generation & Transformation
8.1
Tool Use & Integration
7.4
Perception & Ingestion
4.5
The full twelve-role cohort
RoleTasksAutomation potential
Bookkeeper 8 49.5%
Financial Controller 10 40.5%
FP&A Manager 9 40.5%
Accounts Payable Manager 9 39.6%
Statutory Reporting Manager 9 39.1%
Internal Auditor 9 36.8%
Management Accountant 8 36.0%
Audit Manager 9 35.8%
Treasury Accountant 9 35.7%
Senior Auditor 7 35.6%
Forensic Accountant 8 34.7%
Tax Manager 8 33.5%
Putting AI to work

From potential to practice

Automation potential is an opening, not an instruction. The workflows below are worked examples of how a person and an AI model collaborate on a real task across three loops, with the person leading throughout.

Each workflow follows the same three loops: frame and shape the inputs, execute the heavy lifting, then synthesise and package the result. Every loop carries a copy-ready prompt and a review checkpoint before the work moves on. Three roles are shown below. The same structure was produced for all twelve.

Bookkeeper

Three workflows, ready to adopt

Process payroll

Loop 1: Frame & shape inputs

Define the payrun scope, intake all employee inputs, and confirm rate/deduction rules before any calculation.

The User brings the pay period dates, employee changes (new hires, terminations, rate changes), timesheets, and any one-off adjustments. The LLM is used to assemble a structured payrun input pack, flag missing or inconsistent data, and confirm the rules and rates to apply. Iterate until the input pack is complete. Move on once all inputs are validated.

Prompt · Payrun Input Assembly & Validation
Context:
The User is preparing inputs for a payrun. This prompt assembles the employee-by-employee input set, validates against prior periods, and surfaces exceptions for User attention before calculations begin. No gross-to-net calculations occur in this loop. User Provides:
- Pay period start/end dates and pay date
- Employee roster snapshot: anonymised ID | Status | Pay Type (salary/hourly) | Base Rate | Standard Hours | Department
- New hires and terminations during period (with effective dates)
- Rate or benefit changes effective this period
- Timesheet summary per employee: regular hours, overtime, leave types
- One-off adjustments (bonuses, retro pay, reimbursements, prior-period corrections)
- Benefit elections and voluntary deduction list
- Applicable tax jurisdictions
- Prior payrun totals for comparison Ask:
Produce a validated payrun input pack: per-employee input lines, exception flags (missing timesheet, unusual hours, mid-period change, terminated but still active), and a comparison to the prior payrun highlighting material movements. Return Format:
1. Payrun Input Table, EmpID | Status | Pay Type | Rate | Reg Hrs | OT Hrs | Leave Hrs | One-off Adj | Notes
2. Exception Report, EmpID | Issue | Suggested Resolution
3. New Hires / Terminations Summary
4. Period-over-Period Comparison, total headcount, total hours, total one-offs, % change
5. Open Questions for User, numbered list
6. Confirmed Rules to Apply, restated (OT policy, leave accrual, deduction order) Constraints / Quality Bar:
- Use anonymised employee identifiers
- Do not compute gross or net pay
- Flag any employee with no timesheet but active status
Review before proceeding
  • Roster reflects all hires, terminations, and rate changes
  • All active employees have timesheet or salary basis data
  • Exceptions resolved or escalated
  • One-off adjustments authorised
  • Rules and rates confirmed for the period
Outputs
  • Validated payrun input pack
  • Exception log
  • Confirmed rules and rates
Loop 2: Execute heavy lifting

Calculate gross-to-net pay per employee, reconcile to payroll controls, and prepare draft remittances and filings.

The User brings the validated input pack and applicable tax/deduction tables. The LLM is used to compute gross wages, statutory and voluntary deductions, employer contributions, and net pay; reconcile totals against controls; and draft remittance and filing schedules. Iterate to resolve calculation exceptions and tie out totals. Return to Loop 1 if input errors are discovered; otherwise stay until calculations and drafts are review-ready.

Prompt · Gross-to-Net Calculation & Control Reconciliation
Context:
The User needs per-employee gross-to-net calculations and a reconciliation of totals to payroll control accounts. This prompt performs the core payroll math, flags exceptions, and prepares the basis for remittances and journal entries. User Provides:
- Validated payrun input pack from Loop 1
- Applicable tax tables / rates (income tax, social security/CPP, unemployment, other statutory)
- Voluntary deduction rules (pension, benefits, garnishments) with priority order
- Employer contribution rates
- Payroll control account balances or expected totals
- Year-to-date balances per employee, if relevant for caps/thresholds Ask:
Compute gross wages, each deduction component, employer contributions, and net pay per employee. Reconcile aggregate totals to payroll controls. Identify any employee whose net pay or deduction profile deviates materially from expectation or prior period. Return Format:
1. Gross-to-Net Table, EmpID | Gross | Tax | Social/CPP | Unemployment | Other Statutory | Voluntary Deductions | Net Pay | Employer Contributions
2. Period Totals & Control Reconciliation, line by line tie-out with variances
3. Outlier / Exception Report, EmpID | Metric | Period Value | Prior Period | Variance | Likely Cause
4. Cap / Threshold Hits, employees crossing YTD limits
5. Items Requiring User Decision, bullets Constraints / Quality Bar:
- Show calculation logic for any non-standard line
- Preserve currency and rounding rules consistently
- Use anonymised IDs only
Prompt · Remittance & Filing Draft Preparation
Context:
With calculations validated, the User needs draft government remittances and statutory filings ready for review and submission. This prompt assembles the amounts and schedules required for the period. User Provides:
- Validated gross-to-net outputs from prior prompt
- Applicable remittance types and deadlines (tax withholdings, social security/CPP, unemployment, other)
- Year-end filing requirements if relevant
- Submission method per remittance (portal, file format)
- Prior remittance amounts for comparison Ask:
Produce a remittance schedule with amounts, due dates, and submission methods per item. Draft the content of each filing (in a format the User can transcribe or upload). Compare each to the prior period and flag material movements. Return Format:
1. Remittance Schedule Table, Remittance Type | Amount | Due Date | Submission Method | Status
2. Filing Drafts, one block per filing, showing each required field and value
3. Period-over-Period Variance Commentary
4. Risks / Deadline Flags, items at risk of being missed
5. Review Checklist for User Constraints / Quality Bar:
- Mark all outputs as 'draft – pending User review'
- Do not submit anything
- Tie remittance totals back to the gross-to-net table
Review before proceeding
  • Per-employee calculations spot-checked and reasonable
  • Totals reconcile to payroll controls
  • Outliers explained or corrected
  • Remittance amounts tie back to calculations
  • Filing drafts reviewed for accuracy and completeness
Outputs
  • Gross-to-net register
  • Control reconciliation
  • Draft remittance schedule and filings
Loop 3: Synthesize & package

Post journal entries, prepare distribution artefacts (payslips, payment instructions), and produce the payroll close package for filing and handoff.

The User brings approved calculations, remittances, and any final adjustments. The LLM is used to draft journal entries, generate payslip content per employee, structure the direct deposit / cheque instruction list, and assemble a payroll close summary for management or external providers. Iterate on clarity and completeness. Return to Loop 2 if posting reveals reconciliation gaps; otherwise finalise and file.

Prompt · Payroll Posting, Distribution & Close Package
Context:
The User needs to post payroll to the GL, distribute pay and payslips, and produce a close package suitable for filing and handoff to management or an external provider. This prompt synthesises validated outputs into final-form artefacts. User Provides:
- Approved gross-to-net register and remittance schedule from Loop 2
- GL account mapping for payroll (wages, taxes payable, benefits payable, cash, employer contributions)
- Payment method per employee (direct deposit, cheque)
- Payslip content requirements / standard fields
- Audience for close summary (internal management, external accountant/provider)
- Filing convention for payroll workpapers Ask:
Draft the payroll journal entries, produce a payment instruction list, generate payslip content per employee, and write a concise payroll close summary including totals, exceptions handled, remittances scheduled, and items for management attention. Return Format:
1. Payroll Journal Entries, Dr/Cr lines per account with narratives, totals reconciling to register
2. Payment Instruction List, EmpID | Method | Net Amount | Bank Ref (masked) | Status
3. Payslip Content, per-employee block with standard fields (earnings, deductions, net, YTD)
4. Payroll Close Summary, period, headcount, gross, net, employer cost, remittances due, exceptions handled, items for attention
5. Filing Index, list of artefacts and references Constraints / Quality Bar:
- Mask bank details and use anonymised IDs
- Ensure JE totals reconcile to the gross-to-net register
- Keep close summary concise and audience-appropriate
- Recommendations should be non-decisional and clearly labelled
Review before proceeding
  • Journal entries balance and tie to the register
  • Payment instructions match approved net amounts
  • Payslip content is accurate and complete
  • Close summary reflects the period correctly
  • Workpapers filed with a clear index and approvals captured
Outputs
  • Posted payroll journal entries
  • Payment instruction list
  • Employee payslips
  • Payroll close summary and filed package
Open the other two Bookkeeper workflows

Reconcile general ledger accounts

Loop 1: Frame & shape inputs

Define the reconciliation scope, identify required source documents, and confirm assumptions before any matching or analysis begins.

The User typically brings the reconciliation period, a list of GL accounts in scope (e.g., cash, bank, prepaids, accruals), and awareness of any unusual events during the period. The LLM is used to translate this into a structured scoping brief, a source-document checklist, and a list of clarifying questions or known risk areas. Iterate until the scope, inputs, and assumptions are complete and unambiguous. Move on once the checklist and assumptions are confirmed by the User.

Prompt · Reconciliation Scoping Brief
Context:
The User is a bookkeeper preparing to reconcile general ledger accounts (typically including cash, bank accounts, prepaid expenses, and accruals) for a defined accounting period. This prompt initiates the workflow by producing a clear scoping brief and an input checklist so that downstream matching and analysis are based on a complete, well-defined dataset. The brief should reflect standard reconciliation practice: tying GL balances to independent sources, isolating timing differences, and surfacing items that may require adjusting entries. Do not perform any matching or analysis yet. User Provides:
- Reconciliation period (start and end dates)
- List of GL accounts in scope and their current ending balances (summary level)
- Entity / business unit name (anonymised label is fine)
- Known events during the period that could affect reconciliation (e.g., new bank account, system migration, large one-off transactions)
- Any prior-period unresolved reconciling items carried forward (brief description)
- Approval thresholds or materiality guidance, if known Ask:
Produce a structured reconciliation scoping brief that (1) restates the scope and objectives, (2) lists the exact source documents and extracts required for each in-scope account, (3) identifies likely risk areas and reconciling-item categories to look for, (4) flags missing inputs or ambiguous assumptions, and (5) proposes a sequencing order for reconciling the accounts. Return Format:
1. Scope Summary (period, accounts, entity, materiality)
2. Source Document Checklist, table with columns: Account | Required Document | Source Type (bank, vendor, schedule, GL extract, prior workpaper) | Status placeholder
3. Risk Areas & Expected Reconciling Items, bullets per account
4. Open Questions / Assumptions to Confirm, numbered list
5. Proposed Reconciliation Sequence, ordered list with rationale Constraints / Quality Bar:
- Be specific to the accounts named; avoid generic checklists
- Explicitly mark any assumption made due to missing input
- Keep the brief to one page in length
Review before proceeding
  • Scope, period, and account list accurately reflect the engagement
  • Source document checklist is complete and each item is obtainable
  • Open questions and assumptions have been answered or accepted
  • Prior-period carryforward items are accounted for in the plan
  • Sequencing makes sense given dependencies (e.g., bank before cash GL)
Outputs
  • Reconciliation Scoping Brief
  • Source Document Checklist
  • Confirmed assumptions and open-question log
Loop 2: Execute heavy lifting

Perform the matching, variance identification, and adjusting-entry drafting for each in-scope account using the validated scope and inputs.

The User brings sanitised extracts and summaries of the GL, bank statements, prepaid/accrual schedules, and prior workpapers. The LLM is used to perform line-by-line matching, classify reconciling items (timing vs. error vs. omission), compute prepaid amortisation and accrual movements, and draft adjusting journal entries with explanations. Iterate per account, refine matches, investigate outliers, and re-run when corrections are made. Return to Loop 1 if the scope, balances, or inputs prove materially incorrect; otherwise stay in this loop until each account has a clean reconciling-item list and proposed adjustments.

Prompt · Bank Reconciliation Matching & Variance Analysis
Context:
The User is reconciling one or more bank accounts to the general ledger for the period defined in the validated scoping brief. This prompt performs the core matching exercise: tying bank statement activity to GL cash entries, isolating timing differences (outstanding cheques, deposits in transit, unposted bank fees/interest), and flagging unmatched or suspicious items. Output must support drafting adjusting entries and a reconciliation workpaper. User Provides:
- Validated scoping brief and checklist from Loop 1
- Bank statement transaction extract for the period (date, description, amount, debit/credit), anonymised if needed
- GL cash account transaction extract for the period (date, description, amount, debit/credit)
- Opening and closing balances per bank and per GL
- Prior-period outstanding items still expected to clear Ask:
Match bank transactions to GL entries one-to-one or many-to-one where appropriate. Identify and classify every unmatched item. Compute the reconciled balance and the variance to GL. Propose adjusting journal entries for items that are GL-side corrections, and list bank-side items as outstanding without entries. Return Format:
1. Reconciliation Summary Table, Opening Balance (Bank) | + Deposits | − Withdrawals | Closing Balance (Bank) | Closing Balance (GL) | Variance
2. Matched Items, count and total value (no need to list)
3. Unmatched / Reconciling Items Table, Item | Date | Amount | Side (Bank/GL) | Classification (Timing, Error, Omission, Unknown) | Proposed Action
4. Proposed Adjusting Journal Entries, Dr/Cr lines with account, amount, and one-line narrative
5. Items Requiring User Investigation, bullet list with reason Constraints / Quality Bar:
- Do not invent transactions; mark uncertain matches as 'Unknown'
- Preserve original amounts and signs
- Flag any item above the materiality threshold from Loop 1
Prompt · Prepaid & Accrual Schedule Reconciliation
Context:
The User is reconciling prepaid expense and accrual accounts to their supporting schedules and computing period movements. This prompt compares schedule balances to GL, calculates amortisation or accrual for the period, and drafts adjusting entries. Output should align with the scoping brief and feed the final reconciliation package. User Provides:
- Validated scoping brief from Loop 1
- Prepaid expense schedule (item, vendor, total cost, start date, end date, monthly amortisation, remaining balance)
- Accrual schedule (item, basis, expected amount, period covered)
- GL ending balances for each prepaid and accrual account
- Any new prepaids added or accruals released during the period Ask:
For each schedule line, compute the expected period amortisation or accrual movement, compare the schedule's ending balance to the GL balance, identify variances, and draft adjusting journal entries to align the GL. Return Format:
1. Prepaid Reconciliation Table, Item | Opening | + Additions | − Amortisation | Expected Closing | GL Closing | Variance | Notes
2. Accrual Reconciliation Table, Item | Opening | + New Accrual | − Release/Payment | Expected Closing | GL Closing | Variance | Notes
3. Proposed Adjusting Journal Entries, Dr/Cr lines per variance
4. Items Requiring User Judgement, e.g., contract changes, partial-period items
5. Confirmation of totals tying to GL after proposed entries Constraints / Quality Bar:
- Show calculation logic for any non-trivial amortisation
- Do not propose entries below the materiality threshold unless flagged as required
- Mark estimates clearly
Review before proceeding
  • Every unmatched item has a classification and proposed action
  • Proposed adjusting entries are supportable and within authority limits
  • Variances after proposed entries are zero or explained
  • Items flagged for investigation have been resolved or escalated
  • Schedules tie to GL after entries are applied
Outputs
  • Bank reconciliation workpaper with matched and unmatched items
  • Prepaid and accrual reconciliation tables
  • Draft adjusting journal entries with narratives
  • Open investigation list
Loop 3: Synthesize & package

Produce a reviewer-ready reconciliation summary, approval-ready entry list, and filed workpaper package.

The User brings the validated reconciliation tables, approved adjusting entries, and any commentary on unresolved items. The LLM is used to draft a concise reconciliation summary memo with variance explanations, an approval cover sheet, and a workpaper index for filing. Iterate on tone, completeness, and clarity for the approver. Return to Loop 2 if new variances surface during write-up; otherwise finalise and file.

Prompt · Reconciliation Summary & Approval Package
Context:
The User has completed account-level reconciliations and now needs a packaged summary suitable for management/approver review and audit-trail filing. This prompt converts validated workpapers into a clean, structured summary with variance explanations and a list of entries requiring approval. User Provides:
- Validated reconciliation tables from Loop 2 (bank, prepaid, accrual)
- Final list of proposed adjusting journal entries
- Any unresolved items with planned next steps
- Approver name/role and approval thresholds
- Filing location / workpaper reference convention Ask:
Draft a reconciliation summary memo for the approver, including an executive overview, account-by-account variance commentary, a consolidated list of adjusting entries requiring approval, unresolved items with owners and due dates, and a workpaper index for filing. Return Format:
1. Executive Overview (3–5 sentences: period, accounts reconciled, overall status, total adjustments)
2. Account-by-Account Summary, table: Account | GL Balance | Reconciled Balance | Adjustments Posted | Status (Clean / Adjusted / Open)
3. Variance Explanations, short paragraph per material variance
4. Adjusting Entries for Approval, table: Entry # | Account(s) | Dr | Cr | Narrative | Reason
5. Unresolved Items & Next Steps, Item | Owner | Target Resolution Date
6. Workpaper Index, list of files/references for filing Constraints / Quality Bar:
- Neutral, factual tone suitable for an approver
- No new analysis introduced; only synthesise validated inputs
- Highlight any item above materiality threshold
Review before proceeding
  • Summary accurately reflects the reconciled position and adjustments
  • All entries above approval threshold are listed for sign-off
  • Unresolved items have named owners and dates
  • Workpaper index matches actual files saved
  • Approver has signed off before GL is updated
Outputs
  • Reconciliation Summary Memo
  • Approval-ready adjusting entries list
  • Workpaper index and filed package

Handle accounts payable

Loop 1: Frame & shape inputs

Establish the AP batch scope, intake the invoice population, and confirm coding/approval rules before processing begins.

The User brings the AP batch period or run cut-off, the list of invoices received (in any format), and any known special cases (rush payments, disputed invoices, new vendors). The LLM is used to normalise the invoice intake into a structured queue, surface missing fields, and confirm coding/approval rules to apply. Iterate until the queue is complete and ready for verification. Move on once intake is clean and rules are confirmed.

Prompt · AP Batch Intake & Normalisation
Context:
The User is preparing an accounts payable batch for processing. This prompt structures the incoming invoice population into a normalised queue with extracted key fields, flags missing information, and confirms the coding and approval rules that will govern the rest of the workflow. No payments or postings are made in this loop. User Provides:
- Batch identifier and cut-off date
- List of invoices received with available fields (vendor name, invoice number, date, amount, currency, due date, PO reference, description), anonymised if needed
- Known special cases (rush payments, disputes, new vendors, credit memos)
- Current chart-of-accounts categories typically used for AP coding
- Approval matrix (threshold tiers and approvers, by role)
- Vendor master snapshot or list of known vendors Ask:
Normalise the invoice list into a standard queue, identify duplicates and likely matches to existing vendors, flag invoices with missing or inconsistent fields, and produce a checklist of items needing User attention before verification begins. Return Format:
1. Normalised Invoice Queue, table: Line# | Vendor | Invoice# | Date | Due Date | Amount | Currency | PO Ref | Description | Vendor Match Status | Flags
2. Duplicates / Possible Duplicates, list with rationale
3. Missing Field Report, Line# | Missing/Inconsistent Field | Suggested Action
4. Special Cases Identified, list with handling note
5. Confirmed Coding & Approval Rules, short restatement
6. Open Questions for User, numbered list Constraints / Quality Bar:
- Do not assign GL codes yet
- Mark any inferred field as 'inferred' and show source
- Preserve original amounts and currencies
Review before proceeding
  • Invoice queue is complete and duplicates resolved
  • Missing fields have been filled or flagged with owners
  • Special cases have agreed handling
  • Approval matrix and coding rules are correct for this batch
  • New vendors are routed for setup before processing
Outputs
  • Normalised invoice queue
  • Exception/missing-field log
  • Confirmed coding and approval rules
Loop 2: Execute heavy lifting

Verify, code, three-way match, and prepare payment-ready entries for the validated invoice queue.

The User brings the validated queue plus PO and receiving data where applicable. The LLM is used to perform three-way matching, propose GL coding per invoice, determine approval routing per threshold, and draft the journal entries and payment schedule. Iterate per invoice batch, resolve mismatches, override coding suggestions, and re-run as needed. Return to Loop 1 if intake errors are discovered; otherwise stay until the batch is approval- and payment-ready.

Prompt · Three-Way Match & Coding Proposal
Context:
The User needs to verify invoices against POs and receiving reports, propose GL coding, and route for approval. This prompt performs the matching, suggests codes based on description and prior patterns, and assigns approval tiers using the confirmed matrix. User Provides:
- Validated invoice queue from Loop 1
- PO extract: PO# | Vendor | Line items | Quantity | Unit Price | Total
- Receiving report extract: PO# | Quantity Received | Date
- Chart of accounts (relevant subset)
- Approval matrix from Loop 1
- Recent historical coding examples (vendor → account), if available Ask:
For each invoice, perform a three-way match (invoice vs PO vs receiving), propose a GL code with rationale, calculate any variance, assign the approval tier, and flag exceptions. Return Format:
1. Matched Invoices Table, Line# | Vendor | Invoice# | Amount | PO# | Match Status (Full/Partial/None) | Proposed GL Code | Coding Rationale | Approval Tier | Notes
2. Exceptions Table, Line# | Issue Type (Price variance, Qty variance, No PO, No Receipt, Duplicate) | Detail | Suggested Resolution
3. Approval Routing Summary, Approver | # of invoices | Total value
4. Items Requiring User Decision, bullets Constraints / Quality Bar:
- Tolerance for variance: apply User-confirmed threshold; default to flagging anything >2% or above absolute threshold
- Mark inferred coding clearly
- Do not approve on behalf of approvers
Prompt · Payment Schedule & Journal Entry Draft
Context:
After approvals are obtained, the User needs a payment batch and corresponding journal entries. This prompt builds the payment schedule by method and due date and drafts the AP journal postings. User Provides:
- Approved invoice list from previous prompt (with approval status confirmed)
- Available payment methods per vendor (ACH, cheque, credit card)
- Cash position / payment run constraints (e.g., max batch value)
- Early payment discount terms where applicable
- Currency and FX rate guidance if multi-currency Ask:
Group approved invoices into a payment batch optimised for due date and method, identify any early-payment-discount opportunities, draft the AP and cash journal entries, and produce a payment instruction list. Return Format:
1. Payment Batch Table, Payment# | Vendor | Method | Amount | Currency | Due Date | Discount Captured (Y/N) | Invoice Refs
2. Discount Opportunities, list with savings and required pay-by date
3. Draft Journal Entries, for invoice booking and for payment, with Dr/Cr lines and narratives
4. Batch Totals, by method and by currency
5. Items Held / Excluded, with reason Constraints / Quality Bar:
- Do not execute payments; produce instructions only
- Show calculation for any discount
- Highlight any payment exceeding cash-position constraint
Review before proceeding
  • Three-way match exceptions resolved or formally accepted
  • GL coding reviewed and overrides applied where needed
  • Approvals obtained at the correct tier for every invoice
  • Payment batch totals tie to approved invoice totals
  • Discount opportunities consciously taken or declined
Outputs
  • Coded and matched invoice register
  • Approval-routing record
  • Payment batch instructions
  • Draft journal entries
Loop 3: Synthesize & package

Reconcile AP to vendor statements, produce the AP aging and management report, and package the batch for handoff and filing.

The User brings the executed payment confirmations, posted entries, and vendor statements. The LLM is used to reconcile the AP subledger to vendor statements, surface aging and concentration insights, and draft a management-ready AP report. Iterate on report clarity and discrepancy commentary. Return to Loop 2 if reconciliation reveals processing errors; otherwise finalise and file.

Prompt · AP Reconciliation & Management Report
Context:
The User needs to reconcile the AP subledger to vendor statements, generate an AP aging report, and produce a concise management report for the period. This prompt synthesises validated batch outputs into a reviewer-ready deliverable. User Provides:
- Posted AP subledger balances by vendor (period-end)
- Vendor statements for major vendors (vendor, statement balance, open items)
- Aging buckets policy (e.g., Current, 1–30, 31–60, 61–90, 90+)
- Period totals: invoices processed, payments made, discounts captured, exceptions
- Any unresolved disputes or held items
- Audience and tone for the management report Ask:
Reconcile each major vendor balance to its statement, classify discrepancies, generate the AP aging summary, and draft a management report with KPIs, risks, and recommendations. Return Format:
1. Vendor Reconciliation Table, Vendor | Subledger Balance | Statement Balance | Variance | Likely Cause | Action
2. AP Aging Summary, by bucket, with totals and top vendors per bucket
3. Period KPIs, invoices processed, on-time payment %, discounts captured, exceptions count, average days to pay
4. Risks & Watch Items, bullet list (e.g., concentration, overdue critical vendors)
5. Management Report Draft, short narrative (executive summary + key call-outs)
6. Filing Index, workpaper and document references Constraints / Quality Bar:
- Reconcile to the cent for major vendors; summarise the long tail
- Recommendations should be non-decisional and clearly labelled
- Keep narrative concise and audience-appropriate
Review before proceeding
  • Major vendor balances reconcile or variances are explained
  • AP aging matches subledger totals
  • Management report is accurate and appropriately framed
  • Unresolved items have owners and target dates
  • Batch package is filed with a clear index
Outputs
  • Vendor reconciliation workpaper
  • AP aging report
  • Management report draft
  • Filed batch package with index
Financial Controller

Three workflows, ready to adopt

Lead financial close processes

Loop 1: Frame the close plan, schedule, and ownership

Produce a confirmed close schedule, RACI, and input checklist tailored to the current close cycle.

The User brings the close type (monthly/quarterly/year-end), the target close date, known calendar constraints, the list of entities/units in scope, and any changes since the last close (people, systems, policies). The LLM is used to draft a close schedule, RACI, and stakeholder communication plan, and to surface risks and missing inputs. Iteration is expected as the User adjusts ownership, deadlines, and scope. Move on when the schedule, owners, and required inputs are fully confirmed.

Prompt · Close Schedule, RACI, and Input Checklist Builder
Context:
This prompt initiates a financial close cycle led by a Financial Controller. The objective is to convert known constraints into a workable close schedule, a clear RACI, and a complete input checklist. This is a planning loop, no accounting work occurs here. Outputs are consumed downstream when reconciliations and adjustments are processed. User Provides:
- Close type (monthly / quarterly / year-end) and target close date.
- Entities, business units, and ledgers in scope.
- Known constraints (holidays, system cutovers, audit fieldwork dates, board reporting deadlines).
- Team roster with roles and known absences.
- Changes since last close (new accounts, new policies, systems changes, M&A activity).
- Standard close task list or prior-cycle close calendar (summary form).
- Known pain points or delays from the last close. Ask:
Draft a close schedule with day-by-day milestones, a RACI matrix per task, a stakeholder communication plan, and a complete input checklist mapping each reconciliation/schedule to its owner and source. Highlight risks, single points of failure, and items that should start early. Return Format:
1. Close Schedule (table): Day | Milestone | Task | Owner | Backup | Deadline | Predecessors.
2. RACI Matrix (table): Task | Responsible | Accountable | Consulted | Informed.
3. Input Checklist (table): Item | Source | Owner | Format/Template | Due By.
4. Communication Plan (bullets): cadence, channel, audience, content.
5. Risks and Mitigations (bullets).
6. Open Questions for the User (numbered). Constraints / Quality Bar:
- Sequence tasks to respect predecessors (e.g., reconciliations before adjustments).
- Reflect known absences and capacity constraints.
- Be explicit when ownership is unclear rather than assuming it.
Review before proceeding
  • Schedule respects deadlines, predecessors, and team capacity
  • RACI is confirmed by named owners and backups
  • Input checklist is complete and aligned with the chart of accounts
  • Communication plan is realistic and signed off
  • All open questions are resolved
Outputs
  • Confirmed close schedule
  • Confirmed RACI
  • Input checklist
  • Communication plan
  • Risk log
Loop 2: Process reconciliations, evidence, and adjusting entries

Convert collected reconciliations and supporting schedules into a validated set of adjusting journal entries ready for approval and posting.

The User brings the validated artefacts from Loop 1 plus the collected reconciliations, subledger schedules, intercompany balances, accrual workings, and any flux/variance data. The LLM is used to validate completeness, analyse reconciling items, propose adjusting entries with rationale and standard references, and flag items requiring escalation or auditor attention. Iteration is expected: the User pushes back on entries, supplies missing evidence, and re-runs until each entry is either approved or escalated. Return to Loop 1 if scope or ownership has materially shifted.

Prompt · Reconciliation Completeness and Quality Reviewer
Context:
This prompt reviews the completeness and quality of reconciliations and supporting schedules collected during close. The objective is to surface missing items, weak evidence, and reconciling items requiring adjustment, ahead of journal entry preparation. User Provides:
- Validated artefacts from Loop 1 (input checklist, schedule, RACI).
- Status of each reconciliation (received / outstanding / disputed) with submitter and date.
- Summary content of each reconciliation: account, GL balance, supporting balance, reconciling items with age, description, and amount.
- Supporting evidence types provided (bank confirmations, third-party statements, contracts, calculations).
- Materiality thresholds and any judgement-area focus. Ask:
For each account in scope, assess completeness against the input checklist, evaluate evidence quality, classify reconciling items (timing / error / missing entry / cut-off / estimate), and recommend which items require an adjusting entry, further investigation, or write-off. Flag items likely to draw auditor attention. Return Format:
1. Completeness Status (table): Account | Owner | Status | Evidence Quality (High/Med/Low) | Action.
2. Reconciling Items Analysis (table): Account | Item | Amount | Age | Classification | Recommended Action.
3. Items Requiring Adjusting Entry (list, feeding the next prompt).
4. Auditor Attention Watch List (bullets with rationale).
5. Outstanding Items and Chase List (bullets). Constraints / Quality Bar:
- Apply the User's materiality threshold consistently.
- Distinguish high-confidence classifications from hypotheses.
- Do not invent reconciling item details not present in the inputs.
Prompt · Adjusting Journal Entry Drafter and Rationale Builder
Context:
This prompt drafts adjusting journal entries based on the reconciliation review. Each entry must include a clear rationale, accounting standard reference where relevant, supporting evidence reference, and an approval routing recommendation. Entries will be reviewed and approved by the User before posting. User Provides:
- Validated outputs from the Reconciliation Completeness and Quality Reviewer (especially the Items Requiring Adjusting Entry list).
- Chart of accounts extract (relevant codes).
- Applicable accounting framework (e.g., IFRS, US GAAP) and entity-specific policy notes.
- Approval matrix (thresholds and approvers).
- Period and entity context. Ask:
Draft each adjusting journal entry with debit/credit lines, narrative, standard reference, supporting evidence reference, and routing to the appropriate approver based on threshold. Group entries by category (accruals, reclassifications, corrections, estimates) and provide a brief portfolio summary. Return Format:
1. Journal Entry Schedule (table per entry): JE# | Category | Dr Account | Cr Account | Amount | Narrative | Standard Ref | Evidence Ref | Approver.
2. Portfolio Summary (bullets): total adjustments by category, net P&L impact, balance sheet impact.
3. Entries Requiring Escalation (bullets with rationale).
4. Open Judgement Calls (numbered list for User decision). Constraints / Quality Bar:
- Debits must equal credits per entry.
- Never propose entries that bypass the approval matrix.
- Flag any entry where the supporting evidence is weak or absent.
Review before proceeding
  • All in-scope reconciliations are received and classified
  • Reconciling items are correctly classified and quantified
  • Each adjusting entry has clear rationale, evidence, and approver
  • Net impact on P&L and balance sheet is understood and acceptable
  • Escalations and judgement calls are resolved with documentation
Outputs
  • Reconciliation completeness status report
  • Reconciling items classification
  • Approved adjusting journal entries with audit trail
  • Auditor attention watch list
Loop 3: Finalize statements, checklist, and close package

Produce final financial statements, complete the close checklist, and assemble the archived close package for sign-off and distribution.

The User brings the validated journal entries, post-adjustment trial balance, prior-period comparatives, and any disclosure inputs. The LLM is used to compile draft statements, run variance/flux commentary, complete the close checklist, and prepare an executive close memo. Iteration focuses on commentary quality, disclosure completeness, and presentation. Move on when statements are signed off, the checklist is complete, and the close package is archived.

Prompt · Financial Statement Compiler and Flux Commentary
Context:
This prompt assembles the period's financial statements and flux commentary from the post-adjustment trial balance. The objective is to produce a reviewable draft of the income statement, balance sheet, cash flow statement, and key notes, accompanied by variance commentary suitable for management review. User Provides:
- Validated adjusting entries from Loop 2 and post-adjustment trial balance summary.
- Prior-period and budget comparatives (summary form).
- Standard statement templates and disclosure list.
- Material events or transactions in the period (M&A, impairments, restructurings).
- Applicable accounting framework and any policy elections. Ask:
Compile draft financial statements in the provided format, calculate period-over-period and budget variances at the line-item level, and write concise flux commentary explaining drivers for material movements. Identify disclosures that require new or updated content. Return Format:
1. Draft Income Statement (with PY and Budget columns and % variance).
2. Draft Balance Sheet (with PY column and % variance).
3. Draft Cash Flow Statement (summary).
4. Flux Commentary (per material line: driver, amount, qualitative explanation).
5. Disclosure Update List (bullets: disclosure | reason | proposed content outline).
6. Open Items for User Review (numbered). Constraints / Quality Bar:
- Use the materiality threshold provided to scope commentary.
- Keep commentary factual and driver-based, not speculative.
- Flag any line where the variance cannot be explained from inputs provided.
Prompt · Close Checklist Completion and Executive Close Memo
Context:
This prompt finalises the close by confirming checklist completion, drafting the executive close memo for management and audit sign-off, and structuring the archive package. User Provides:
- Validated draft financial statements and flux commentary from the prior prompt.
- Status of each close checklist item (complete / open / N/A) with owner.
- Sign-off requirements (who must approve what).
- Key themes for the memo (close timeliness, control issues, judgement areas, audit readiness).
- Archive structure or document repository conventions. Ask:
Confirm checklist completeness, draft an executive close memo summarising results, key judgements, control observations, and outstanding items, and produce an archive index mapping each artefact to its location. Return Format:
1. Close Checklist Status (table): Item | Owner | Status | Notes.
2. Executive Close Memo (1–2 pages): Period Summary | Key Results | Material Judgements | Control and Process Observations | Outstanding Items | Recommended Sign-offs.
3. Archive Index (table): Artefact | Owner | Location | Retention.
4. Sign-off Routing (table): Approver | Scope | Deadline. Constraints / Quality Bar:
- The memo must be readable by non-accountants without losing technical accuracy.
- Outstanding items must each have an owner and deadline.
- Do not declare the close complete if any material checklist item is open.
Review before proceeding
  • Financial statements tie to the trial balance and supporting schedules
  • Flux commentary is accurate, driver-based, and within materiality
  • All checklist items are complete or have a documented exception
  • Sign-offs are obtained from the correct approvers
  • Close package is archived per convention
Outputs
  • Final financial statements with flux commentary
  • Completed close checklist
  • Executive close memo
  • Archived close package and index
Open the other two Financial Controller workflows

Direct day-to-day accounting operations

Loop 1: Frame the daily operations run

Establish a clear, prioritised picture of today's accounting workload, exceptions, and decision points before any processing begins.

The User brings a high-level snapshot of inbound items (AP queue counts, AR aging changes, payroll cycle stage, bank feed status, tax deadlines). The LLM is used to structure that snapshot into a prioritised daily run-sheet, surface anomalies, and flag missing inputs. Iteration is expected: the User refines scope, adds missing items, and re-runs until the run-sheet reflects reality. Move on when priorities, exceptions, and required inputs are confirmed.

Prompt · Daily Accounting Run-Sheet Framer
Context:
This prompt supports a Financial Controller preparing the day's accounting operations across AP, AR, payroll, reconciliations, and tax compliance. The objective is to convert a raw snapshot of inbound items into a prioritised run-sheet that identifies exceptions, missing inputs, and decision points before any processing begins. This is the framing step, no transactions are processed here. The output will be consumed by a downstream execution loop. User Provides:
- Date and accounting period stage (e.g., mid-month, pre-close).
- AP snapshot: count of new vendor bills, total value band, any flagged duplicates or missing POs.
- AR snapshot: new customer invoices to issue, overdue invoices in each aging bucket, disputed items.
- Payroll status: cycle stage (data collection / calculation / approval / disbursement), known exceptions (new hires, terminations, variable pay).
- Bank and reconciliation status: unreconciled items count, age, bank feed health.
- Tax and remittance deadlines falling within the next 5 business days.
- Known stakeholder issues or pending approvals.
- Standing thresholds (e.g., payment approval limits, materiality for variance flags). Ask:
Structure the inputs into a prioritised daily run-sheet. Identify which items are routine, which are exceptions requiring judgement, and which are blocked pending missing inputs or approvals. Surface any deadline conflicts and propose a sequencing of work. Return Format:
1. Priority Run-Sheet (table): Item | Category (AP/AR/Payroll/Recon/Tax) | Routine vs Exception | Owner needed | Deadline | Suggested order.
2. Exceptions Requiring Judgement (bullets with brief rationale).
3. Missing Inputs / Blockers (bullets with who to chase).
4. Deadline Conflicts or Risks (bullets).
5. Open Questions for the User (numbered). Constraints / Quality Bar:
- Do not assume access to live systems; reason only from inputs provided.
- Flag any item where information is insufficient rather than guessing.
- Keep exception rationale to one or two sentences.
Review before proceeding
  • Run-sheet priorities align with actual operational urgency and deadlines
  • All known exceptions and blockers are captured
  • Missing inputs have been identified and chased
  • Open questions from the LLM have been answered or accepted as assumptions
Outputs
  • Validated daily run-sheet
  • Exceptions list
  • Blockers / missing inputs list
Loop 2: Execute processing, reconciliations, and exception analysis

Work through the validated run-sheet to produce structured outputs for each AP, AR, payroll, reconciliation, and tax item, with discrepancies isolated for User decision.

The User feeds the validated run-sheet plus item-level details (sanitised invoice data, payroll register summaries, bank vs ledger differences, tax calculation inputs). The LLM is used to validate entries against rules, propose journal coding, identify reconciliation breaks, and analyse discrepancies. Iteration is expected: the User pushes back on suggestions, supplies additional context for ambiguous items, and re-runs until each line is either ready for execution or escalated. Return to Loop 1 if the scope of the day's work has materially changed.

Prompt · AP/AR Validation and Discrepancy Analyser
Context:
This prompt processes vendor bills and customer invoices flagged in the validated daily run-sheet. The objective is to validate each item against standing rules (PO match, tax codes, duplicate check, approval thresholds), propose GL coding, and isolate discrepancies for review. Routine items should be cleared for execution; exceptions should be analysed with recommended next steps. User Provides:
- Validated run-sheet from Loop 1 (relevant AP/AR section).
- Line-level details for each item: vendor/customer, amount, currency, date, reference, PO/contract reference, supporting document type.
- Chart of accounts extract (relevant codes and descriptions).
- Standing rules: approval thresholds, tax code mapping, duplicate detection criteria.
- Known disputes or aging issues for AR items. Ask:
For each item, validate completeness, propose GL coding and tax treatment, flag potential duplicates or PO mismatches, and recommend whether to (a) clear for execution, (b) hold for missing info, or (c) escalate as a discrepancy. For each discrepancy, provide a short root-cause hypothesis and proposed resolution path. Return Format:
Table: Item Ref | Counterparty | Amount | Proposed GL/Tax Coding | Validation Result | Recommendation (Clear / Hold / Escalate) | Notes.
Below the table: Discrepancy Analysis section with one paragraph per escalated item (root cause hypothesis, resolution path, who to involve). Constraints / Quality Bar:
- Never recommend Clear when any required field is missing or ambiguous.
- Coding suggestions must reference the provided chart of accounts.
- Keep notes terse and decision-oriented.
Prompt · Payroll, Reconciliation, and Tax Review Analyser
Context:
This prompt supports review of payroll calculations, account reconciliations, and tax liability positions identified in the validated run-sheet. The objective is to detect calculation errors, reconciling differences, and tax exposure issues, and to recommend adjusting actions. No postings are executed here, the output drives User decisions and downstream entries. User Provides:
- Validated run-sheet from Loop 1 (payroll, reconciliation, and tax sections).
- Payroll summary: gross pay, withholdings, employer contributions, period-over-period delta, exception list.
- Reconciliation inputs: account, GL balance, subledger or bank balance, unreconciled items with age and description.
- Tax inputs: liability accruals, remittance amounts due, prior period comparatives, filing deadlines.
- Materiality thresholds and standing tolerance levels. Ask:
For payroll, identify outliers vs prior period and validate withholding logic against provided rates. For reconciliations, classify each unreconciled item (timing, error, missing entry, fraud risk) and propose clearing entries where applicable. For tax positions, verify liability accruals against remittance amounts and flag any shortfall or over-accrual. Return Format:
1. Payroll Review: table of variances above threshold | likely cause | recommended action.
2. Reconciliation Findings: per-account table, Account | Difference | Classification | Proposed Entry or Action | Confidence.
3. Tax Position Review: table, Tax type | Liability | Remittance due | Variance | Recommendation.
4. Items Requiring Escalation: bullet list with rationale. Constraints / Quality Bar:
- Distinguish clearly between high-confidence findings and hypotheses requiring User verification.
- Do not propose entries above the User's stated approval threshold without flagging them for escalation.
- Surface any pattern suggesting a control weakness.
Review before proceeding
  • Each AP/AR item is correctly classified as Clear, Hold, or Escalate
  • Proposed GL coding and tax treatment have been spot-checked
  • Payroll variances and reconciliation classifications are accepted or corrected
  • All recommended adjusting entries are within approval authority or escalated
  • Escalations have a clear next-step owner
Outputs
  • Validated AP/AR processing list with coding and recommendations
  • Payroll variance and exception findings
  • Reconciliation classification and proposed entries
  • Tax position review with flagged variances
  • Consolidated escalation list
Loop 3: Synthesize daily report and leadership handoff

Package the day's processed activity, exceptions, and decisions into a concise variance and exception report suitable for finance leadership.

The User brings the validated outputs from Loop 2 plus any post-execution updates (payments actually released, escalations resolved). The LLM is used to draft a daily summary report covering throughput, variances, exceptions, and items requiring leadership attention. Iteration is expected on tone, level of detail, and call-out emphasis. Move on when the report is accurate, decision-oriented, and ready to distribute.

Prompt · Daily Operations Summary and Exception Report
Context:
This prompt produces the end-of-day operations summary for finance leadership. The audience is the CFO and senior finance leaders who need a fast read on what was processed, what exceptions arose, what was decided, and what remains outstanding. The tone is concise, factual, and decision-oriented. User Provides:
- Validated outputs from Loop 2 (AP/AR processing list, payroll findings, reconciliation findings, tax review, escalation list).
- Post-execution updates: payments released, receipts applied, entries posted, escalations resolved or still open.
- Any leadership-specific context (e.g., upcoming board meeting, cash position sensitivity).
- Preferred report length and format. Ask:
Draft a daily operations report that summarises throughput, highlights material variances and exceptions, captures decisions made, and lists open items with owners and deadlines. Return Format:
1. Headline (2–3 sentences): state of operations today.
2. Throughput Snapshot (table): AP processed, AR issued, payroll status, reconciliations cleared, tax remittances.
3. Material Variances and Exceptions (bullets with amounts and brief explanation).
4. Decisions Made Today (bullets).
5. Open Items and Escalations (table: Item | Owner | Deadline | Status).
6. Watch List for Tomorrow (bullets). Constraints / Quality Bar:
- Lead with what leadership needs to act on, not what was routine.
- Quantify exceptions wherever amounts are available.
- Keep the full report to one page when printed.
Review before proceeding
  • Throughput numbers reconcile to actual processed activity
  • Material exceptions are accurately characterised
  • Decisions and open items are correctly attributed
  • Tone and length match leadership expectations
Outputs
  • Final daily operations summary report
  • Open items / escalation tracker update

Oversee financial statement preparation

Loop 1: Frame data scope, reporting basis, and audit coordination plan

Confirm the reporting scope, accounting framework, data sources, comparatives, and audit coordination plan before drafting begins.

The User brings the reporting period, entity and consolidation scope, applicable framework, audit timeline, and known issues from prior periods. The LLM is used to produce a data inventory, validation checklist, audit coordination plan, and risk map. Iteration is expected as scope and timelines firm up. Move on when the inventory, plan, and risk map are confirmed.

Prompt · Reporting Scope, Data Inventory, and Audit Plan Framer
Context:
This prompt initiates the financial statement preparation cycle. The objective is to define reporting scope, list and validate required data inputs, and establish an audit coordination plan. This is a framing step, no statements are drafted here. User Provides:
- Reporting period and entity/consolidation scope.
- Applicable accounting framework and any new standards effective this period.
- List of source systems and owners (general ledger, subledgers, treasury, payroll).
- Comparative periods required.
- External audit firm and known audit timeline / milestones.
- Prior-period audit findings, management letter points, or restatements.
- Known accounting judgement areas (e.g., revenue recognition, impairment, leases). Ask:
Produce a data inventory and validation checklist, an audit coordination plan with milestones and PBC (Prepared-By-Client) categories, and a risk map highlighting judgement areas and prior-period concerns to address proactively. Return Format:
1. Data Inventory (table): Data Item | Source | Owner | Format | Validation Check | Due Date.
2. Reporting Basis Confirmation (bullets): framework, scope, key policy elections, new standards applicable.
3. Audit Coordination Plan (table): Milestone | Date | Auditor Lead | Client Lead | Deliverables.
4. Risk Map (table): Area | Risk | Driver | Planned Response.
5. Open Questions for the User (numbered). Constraints / Quality Bar:
- Be explicit about any scope ambiguity.
- Tie each data item to an owner and a validation check.
- Surface new-standard impacts even if not yet quantified.
Review before proceeding
  • Reporting scope and basis are confirmed with leadership
  • Data inventory is complete with owners and due dates
  • Audit coordination plan is aligned with the audit firm
  • Risk map reflects all known judgement and prior-period issues
  • Open questions are resolved or assigned
Outputs
  • Validated data inventory
  • Confirmed reporting basis
  • Audit coordination plan
  • Risk map
Loop 2: Draft statements, validate data, and respond to audit queries

Produce a fully drafted set of financial statements with disclosures, supported by validated data and an organised audit response trail.

The User brings validated artefacts from Loop 1, collected data (general ledger summary, trial balance, reconciliations, intercompany schedules, comparatives), and incoming auditor queries. The LLM is used to validate data consistency, draft statements and notes, perform consolidation logic checks, and draft responses to audit requests. Iteration is expected on technical accuracy, disclosure completeness, and audit query quality. Return to Loop 1 if scope, framework, or comparatives change materially.

Prompt · Data Validation and Consolidation Consistency Checker
Context:
This prompt validates the financial data assembled for statement preparation. The objective is to confirm internal consistency across the trial balance, reconciliations, intercompany schedules, and comparatives, and to surface any anomalies before drafting statements. User Provides:
- Validated artefacts from Loop 1 (data inventory, reporting basis, risk map).
- Trial balance summary by entity and consolidated.
- Reconciliation status by account.
- Intercompany matrix (counterparty balances and eliminations).
- Comparative period balances.
- Known one-off transactions and their treatment. Ask:
Check that trial balances tie to reconciliations, intercompany balances eliminate cleanly, comparatives align with prior published figures, and unusual movements vs comparatives are flagged. Identify data issues that must be resolved before drafting. Return Format:
1. Tie-Out Status (table): Check | Result (Pass/Fail/Review) | Difference | Notes.
2. Intercompany Elimination Analysis (table): Pair | Reported by A | Reported by B | Difference | Action.
3. Comparative Consistency Check (table): Line | Prior Published | Current System | Difference | Explanation Needed.
4. Unusual Movements (bullets with amounts).
5. Blockers Requiring Resolution Before Drafting (numbered). Constraints / Quality Bar:
- Do not proceed to drafting if material blockers remain.
- Quantify all differences.
- Distinguish data errors from real economic movements.
Prompt · Financial Statement and Disclosure Drafter
Context:
This prompt drafts the financial statements and notes once data has been validated. The objective is to produce a complete, framework-compliant draft suitable for management review and audit. User Provides:
- Validated outputs from the data validation prompt.
- Adjusted trial balance and consolidated figures.
- Statement and note templates / prior-period statements as reference.
- Applicable framework and policy notes.
- List of required disclosures including any new-standard impacts.
- Material events, transactions, and judgements to disclose. Ask:
Draft the primary statements (income statement, balance sheet, cash flow, equity movement) and all required notes, applying the framework consistently. Draft new or updated disclosures for material events and new standards. Identify gaps requiring User input. Return Format:
1. Primary Statements (formatted drafts with comparatives).
2. Notes to the Statements (numbered, framework-compliant, with cross-references).
3. New / Updated Disclosures (list with rationale and draft text).
4. Drafting Gaps and Judgement Calls (numbered, for User decision).
5. Cross-Reference Map (table): Statement Line | Supporting Note | Workpaper Reference. Constraints / Quality Bar:
- Apply the stated framework consistently; do not mix frameworks.
- Flag rather than fabricate any number or disclosure not supported by inputs.
- Keep language consistent with the prior-period statements unless a change is justified.
Prompt · Audit Query Response Drafter
Context:
This prompt drafts responses to auditor queries during fieldwork. The objective is to produce clear, evidence-backed responses that minimise back-and-forth and document a defensible position. User Provides:
- Validated artefacts from prior steps (drafts, validation results, risk map).
- Auditor query text and context.
- Relevant supporting evidence (summary or excerpt) and workpaper references.
- Management position on any judgement involved. Ask:
Draft a structured response that restates the query, sets out the management position, references supporting evidence, and identifies any open items needing further work. Flag any query suggesting a potential audit adjustment. Return Format:
1. Query Restatement.
2. Management Position (with framework reference where applicable).
3. Supporting Evidence References.
4. Open Items / Further Work Needed.
5. Potential Adjustment Flag (Yes/No with reasoning). Constraints / Quality Bar:
- Be precise and non-defensive in tone.
- Cite only evidence the User has confirmed exists.
- Surface any query that could lead to a material adjustment.
Review before proceeding
  • Data tie-outs and intercompany eliminations pass
  • Drafts apply the framework consistently and tie to the trial balance
  • Disclosures are complete, including new-standard impacts
  • Audit query responses are evidence-backed and defensible
  • Material judgement calls are documented and approved
Outputs
  • Data validation report
  • Draft financial statements and notes
  • Disclosure register
  • Audit query response log
  • Potential audit adjustment list
Loop 3: Finalize, approve, and distribute

Incorporate audit adjustments, obtain approvals, complete filings, and distribute the final financial statements.

The User brings the agreed audit adjustments, final drafts, approval requirements, and filing/distribution lists. The LLM is used to produce the final statement package, board/management memo, filing-ready outputs, and distribution plan. Iteration focuses on presentation, executive summary clarity, and filing accuracy. Move on when approvals are recorded and distribution is executed.

Prompt · Audit Adjustment Integrator and Final Statement Packager
Context:
This prompt integrates final audit adjustments and produces the final statement package for approval, filing, and distribution. User Provides:
- Validated drafts and audit query log from Loop 2.
- Final list of agreed audit adjustments (and any passed adjustments memo).
- Approval requirements (management, audit committee, board).
- Regulatory filing requirements and formats.
- Distribution list (internal and external) with channels. Ask:
Apply audit adjustments to the drafts, recompute affected lines and disclosures, produce the final statement package, draft a board/audit-committee memo summarising results and key judgements, prepare a passed adjustments summary, and produce a filing and distribution plan. Return Format:
1. Final Statements (revised with audit adjustments applied; change log vs prior draft).
2. Audit Adjustments Applied (table): Adjustment | Dr/Cr | Amount | Reason | Source.
3. Passed Adjustments Summary (table with materiality rationale).
4. Board / Audit Committee Memo (1–2 pages): Period Results | Key Judgements | Audit Outcome | Recommended Approvals.
5. Filing and Distribution Plan (table): Recipient/Filing | Format | Owner | Deadline | Status.
6. Sign-off Routing (table): Approver | Scope | Deadline. Constraints / Quality Bar:
- The change log must reconcile every difference vs the prior draft.
- Passed adjustments rationale must reference materiality explicitly.
- Filing outputs must match required formats; flag any format uncertainties.
Review before proceeding
  • All agreed audit adjustments are applied and reconciled
  • Affected disclosures and totals are recomputed correctly
  • Board / audit committee memo is approved
  • Regulatory filing outputs match required formats
  • Distribution is executed to the correct recipients on time
Outputs
  • Final approved financial statements
  • Audit adjustments log and passed adjustments summary
  • Board / audit committee memo
  • Filing confirmations
  • Distribution record
Tax Manager

Three workflows, ready to adopt

Prepare and review tax returns

Loop 1: Frame & shape inputs

Establish a validated engagement profile, document inventory, and preparation plan for each return before any drafting begins.

The User typically brings a portfolio of entities (C-Corps, S-Corps, Partnerships, HNW Individuals), prior-year returns, and an initial set of client-provided documents. The LLM is used to translate this raw situation into a structured intake checklist, identify gaps, and produce a return-by-return preparation plan. Iteration is expected as the User receives more documents or clarifies entity facts. Move on once the intake list, open items, and preparation plan are confirmed accurate and complete for each return in scope.

Prompt · Engagement Scoping & Intake Checklist Builder
Context
The User is a Tax Manager leading preparation and review of complex federal, state, and local returns for one or more entities (C-Corp, S-Corp, Partnership, or High Net Worth Individual). Before drafting begins, a structured intake plan must be created that captures entity facts, filing footprint, prior-year carryovers, expected elections, and required source documents. This prompt is the first step in a three-loop workflow and exists to convert raw engagement context into a clean preparation plan. Do not perform tax calculations or draft returns at this stage. User Provides
- Entity type and tax year: [ENTITY TYPE + YEAR]
- Summary of entity facts (states of operation, ownership, business activity, major events): [SUMMARY]
- Prior-year return highlights (carryovers, elections, methods, NOLs, credits, K-1 activity): [PRIOR YEAR SUMMARY]
- Known client-provided documents already received (list by type): [DOCUMENTS RECEIVED]
- Known open items, client questions, or unusual transactions: [OPEN ITEMS]
- Filing deadlines and extensions filed: [DEADLINES] Ask
Produce a return-specific intake plan that (1) lists all source documents and data points required to prepare this return, (2) flags missing or ambiguous items, (3) highlights carryovers and elections that must be confirmed, (4) identifies state and local filings triggered by the facts, and (5) lists targeted clarification questions to send to the client or internal accounting team. Return Format
1. Engagement Snapshot (entity, year, jurisdictions, deadlines)
2. Required Documents Checklist (table: Document | Purpose | Status: Received / Missing / Unclear)
3. Carryovers & Elections to Confirm (table: Item | Prior-Year Value | Source | Action)
4. State & Local Filing Footprint (table: Jurisdiction | Filing Type | Trigger | Notes)
5. Open Questions for Client / Internal Team (numbered list)
6. Preparation Plan Outline (sequenced steps, dependencies, review checkpoints) Constraints / Quality Bar
- Be specific to the entity type; do not produce generic checklists.
- Distinguish clearly between confirmed facts and assumptions.
- Flag any item where missing data would block draft preparation.
Prompt · Document Ingestion & Gap Analysis
Context
The User has gathered an initial set of client documents (financial statements, trial balances, prior-year returns, K-1s, payroll reports, tax notices, entity formation documents). This prompt exists to normalize what has been received, cross-check against the intake checklist from the prior prompt, and isolate gaps and inconsistencies before drafting begins. User Provides
- Validated intake checklist from prior prompt: [INTAKE CHECKLIST]
- Summaries or excerpts of documents received (sanitised, no PII beyond what is necessary): [DOCUMENT SUMMARIES]
- Any client commentary or notes accompanying the documents: [CLIENT NOTES] Ask
Reconcile the received documents against the checklist, identify missing or inconsistent items, surface anomalies (e.g., trial balance not tying to financials, K-1 figures inconsistent with capital accounts), and produce a prioritized follow-up list for the client and internal team. Return Format
1. Coverage Matrix (table: Checklist Item | Document Mapped | Coverage: Complete / Partial / Missing | Notes)
2. Inconsistencies & Anomalies (table: Item | Observation | Possible Cause | Suggested Resolution)
3. Prioritized Follow-Up List (P1/P2/P3 with owner suggestion)
4. Readiness Assessment (Ready to Draft / Ready with Caveats / Not Ready) with rationale Constraints / Quality Bar
- Do not infer numbers that are not present; mark them as missing instead.
- Cite which document a finding came from.
Review before proceeding
  • Entity type, tax year, and filing footprint are confirmed correct
  • All required source documents are received or explicitly tracked as open items
  • Prior-year carryovers, methods, and elections have been verified against prior return
  • Open client questions have been sent and material answers received
  • Readiness Assessment is 'Ready to Draft' or 'Ready with Caveats' with caveats accepted
Outputs
  • Validated intake checklist
  • Coverage matrix and gap log
  • Preparation plan outline
  • Open items / clarification log
Loop 2: Execute heavy lifting

Produce a complete, technically reviewed draft return with supporting schedules, identified corrections, and documented planning observations.

The User brings the validated intake, normalized financial data, and the preparation plan from Loop 1. The LLM is used to draft return line items and schedules from structured inputs, then to perform a layered technical review that surfaces inconsistencies, missed elections, uncertain tax positions, and provision impacts. Iteration is expected as numbers are refined, elections are decided, and review notes are cleared. Move on once the draft return, schedules, and review notes are at a state the User would sign off on technically.

Prompt · Draft Return & Supporting Schedules Generator
Context
The User is preparing a draft return for the entity identified in Loop 1 (C-Corp, S-Corp, Partnership, or HNW Individual). This prompt generates a structured draft of return line items and supporting schedules from the validated inputs. It does not replace tax software computations; it produces a reviewable structured draft and a working schedule set that can be reconciled to software output. User Provides
- Validated intake and gap analysis from Loop 1: [LOOP 1 ARTEFACTS]
- Trial balance / income statement / balance sheet summary: [FINANCIALS]
- Book-to-tax adjustments identified so far: [M-1 / M-3 ITEMS]
- Apportionment data by state (if applicable): [APPORTIONMENT DATA]
- Owner / shareholder / partner data (allocations, basis, distributions): [OWNER DATA]
- Credits, incentives, and elections being claimed: [CREDITS & ELECTIONS]
- Prior-year carryovers (NOLs, capital losses, credits, AMT, etc.): [CARRYOVERS] Ask
Draft the structured federal return line items and supporting schedules for the specified entity, including book-to-tax reconciliation, state apportionment workings, K-1 / K-3 allocations where applicable, credit computations, and a list of elections and disclosures to attach. Show calculations transparently. Return Format
1. Federal Return Draft (table: Line | Description | Amount | Source / Calculation)
2. Book-to-Tax Reconciliation (M-1 / M-3 style schedule)
3. State & Local Schedules (per jurisdiction: apportionment, modifications, tax computation)
4. Owner Allocations (K-1 / Schedule K data per owner, where applicable)
5. Credits & Elections Workpaper (citation to authority, computation, supporting docs)
6. Disclosure & Attachment List
7. Assumptions & Open Items used in this draft Constraints / Quality Bar
- Show every figure's source or calculation path.
- Where authority is cited, name the Code section, regulation, or form instruction.
- Flag any computation that depends on an unresolved open item.
Prompt · Technical Review & Issues Log
Context
The User has a draft return and supporting schedules and now requires a meticulous technical review covering accuracy, internal consistency, cross-entity impacts, provision implications, uncertain tax positions, and filing compliance. This prompt produces a structured issues log that the User will resolve before finalization. User Provides
- Draft return and schedules from prior prompt: [DRAFT RETURN ARTEFACTS]
- Related-entity returns or tax provision data (if relevant): [RELATED RETURNS / PROVISION]
- Known sensitive positions or prior IRS / state correspondence: [SENSITIVE HISTORY]
- Materiality thresholds and review depth required: [MATERIALITY] Ask
Perform a layered technical review and produce a prioritized issues log covering: arithmetic and tie-out checks, internal consistency (book/tax, federal/state, K-1/capital accounts), missed elections or credits, treatment of unusual transactions, uncertain tax positions, tax provision and deferred tax impacts, and filing/signature/disclosure compliance. Identify planning opportunities surfaced by the data. Return Format
1. Review Summary (overall readiness, key risks)
2. Issues Log (table: ID | Category | Severity: High/Med/Low | Description | Reference | Recommended Action)
3. Cross-Entity / Provision Impacts
4. Uncertain Tax Positions to Consider (with rationale)
5. Planning Opportunities Identified (with brief description and savings indicator)
6. Compliance & Filing Checklist (signatures, e-file readiness, disclosures, statements) Constraints / Quality Bar
- Cite the specific schedule, line, or workpaper for each issue.
- Distinguish 'must fix before filing' from 'consider for planning' clearly.
- Do not introduce new tax positions without citing authority.
Review before proceeding
  • All High-severity issues in the issues log have been resolved or consciously accepted
  • Draft figures tie to tax software output and supporting workpapers
  • Elections, disclosures, and statements are confirmed and attached
  • Cross-entity and tax provision impacts are reconciled
  • Uncertain tax positions have been reviewed and dispositioned
  • Planning opportunities have been logged for client discussion
Outputs
  • Reviewed draft return and schedules
  • Technical review issues log with dispositions
  • Elections & disclosures workpaper
  • Planning opportunities list
Loop 3: Synthesize & package

Produce client-ready deliverables, filing package, and archived working papers with clear explanations and action items.

The User brings the reviewed draft and cleared issues log from Loop 2. The LLM is used to package the return into a client deliverable (cover letter, summary of results, payment instructions, planning observations) and a filing package checklist for e-file or paper filing. Iteration is expected as tone, level of detail, and emphasis are adjusted for the client. Move on once deliverables are approved by the User and the filing and archive checklist is complete.

Prompt · Client Deliverable & Filing Package Builder
Context
The User has a finalized, technically reviewed return and now needs a client-facing deliverable package and an internal filing/archival checklist. This prompt synthesizes the Loop 2 outputs into a clear, professional client communication and a complete filing package, without introducing new tax positions. User Provides
- Reviewed return and resolved issues log from Loop 2: [LOOP 2 ARTEFACTS]
- Client profile and preferred tone (e.g., concise executive, detail-oriented HNW, finance team): [CLIENT PROFILE]
- Payment, refund, and estimated tax positions: [PAYMENT POSITIONS]
- Planning observations to highlight to the client: [PLANNING ITEMS]
- Filing channels (e-file, paper, jurisdiction-specific): [FILING CHANNELS] Ask
Produce (1) a client-facing cover letter and results summary, (2) a payment/refund/estimated-tax instruction sheet, (3) a planning observations memo tailored to the client's profile, and (4) an internal filing and archive checklist covering signatures, e-file authorizations, attachments, transmittal, payment confirmation, and workpaper archival. Return Format
1. Client Cover Letter (professional, plain-language, no jargon overload)
2. Results Summary (table: Jurisdiction | Tax | Payments/Credits | Balance Due / Refund | Due Date)
3. Payment & Estimated Tax Instructions (per jurisdiction, with method and timing)
4. Planning Observations Memo (bullet list with rationale and suggested next steps)
5. Internal Filing & Archive Checklist (table: Item | Owner | Status | Due)
6. Suggested Client Follow-Up Items (questions, planning meetings, future-year actions) Constraints / Quality Bar
- Match tone to the client profile provided.
- Do not introduce positions or numbers not present in the reviewed return.
- Keep the cover letter concise; place detail in the appendices/checklists.
Review before proceeding
  • Cover letter and summary accurately reflect the final return
  • Payment and estimated tax instructions are correct per jurisdiction
  • Planning observations are appropriate to share and free of speculative positions
  • Internal filing and archive checklist is complete and assigned
  • Signatures, e-file authorizations, and confirmations are accounted for
Outputs
  • Client cover letter and results summary
  • Payment and estimated tax instruction sheet
  • Planning observations memo
  • Internal filing and archive checklist
Open the other two Tax Manager workflows

Manage client relationships

Loop 1: Frame & shape inputs

Establish a clear client account profile, service map, and information inventory that drives all downstream work.

The User typically brings an engagement letter, prior-year files, and a mix of recent client documents and emails. The LLM is used to organize this into a client account profile, identify upcoming deliverables and deadlines, and surface missing inputs or ambiguities. Iteration is expected as documents arrive and the client clarifies expectations. Move on once the account profile, deliverable map, and intake gaps are confirmed.

Prompt · Client Account Profile & Service Map
Context
The User is the primary point of contact for a tax client and needs a structured account profile that captures the client's entities, deliverables, deadlines, contacts, and service expectations. This prompt exists to convert scattered engagement context into a single working profile that anchors all subsequent client interactions. User Provides
- Client name / type and entity structure summary: [CLIENT SUMMARY]
- Engagement letter scope and fee arrangement summary: [ENGAGEMENT SCOPE]
- Prior-year deliverables and notable issues: [PRIOR-YEAR HIGHLIGHTS]
- Known upcoming deadlines and deliverables: [DEADLINES]
- Key client contacts and their roles: [CONTACTS]
- Communication preferences and sensitivities: [PREFERENCES] Ask
Build a complete client account profile and service map that captures scope, contacts, deliverables, deadlines, recurring touchpoints, and known risk areas. Identify any ambiguities in scope or gaps in expectations. Return Format
1. Client Snapshot (entities, jurisdictions, fiscal calendar)
2. Service & Deliverable Map (table: Deliverable | Frequency | Owner | Next Due Date)
3. Contact Directory (table: Name | Role | Topics | Preferred Channel)
4. Recurring Touchpoints (cadence and agenda themes)
5. Known Risk & Watch Areas (positions, prior issues, sensitive topics)
6. Scope Ambiguities & Clarifications Needed Constraints / Quality Bar
- Stay within the engagement scope provided; do not invent services.
- Flag assumptions explicitly.
Prompt · Information Intake & Validation Plan
Context
The User needs to assemble and validate client information (financial statements, prior returns, supporting documents) required to advance the account. This prompt produces an intake checklist and validation plan so that incomplete or inconsistent data is surfaced early. User Provides
- Validated client account profile from prior prompt: [ACCOUNT PROFILE]
- Documents already received (sanitised summaries): [DOCUMENTS RECEIVED]
- Current open client requests or questions: [OPEN REQUESTS] Ask
Cross-reference received documents against the service map, list missing items, propose reconciliation checks (e.g., financials vs. prior return, payroll vs. GL, K-1 ties), and draft a client outreach plan to close gaps. Return Format
1. Document Coverage Matrix (Required | Received | Status | Notes)
2. Validation & Reconciliation Checks to Run
3. Gap-Closure Outreach Plan (priority, recipient, suggested message theme)
4. Readiness Assessment for next deliverable(s) Constraints / Quality Bar
- Do not infer figures; mark missing items clearly.
- Group asks to the client efficiently to limit back-and-forth.
Review before proceeding
  • Client account profile accurately reflects engagement scope and contacts
  • Deliverables and deadlines are complete and confirmed
  • Document intake matrix reflects current state of receipts
  • Outreach plan is prioritized and ready to execute
  • No critical scope ambiguity remains unresolved
Outputs
  • Client account profile and service map
  • Document coverage matrix
  • Client outreach plan
  • Open items log
Loop 2: Execute heavy lifting

Validate client financials, prepare draft deliverables, and resolve client questions to a near-final state.

The User brings the validated account profile, intake matrix, and current client data from Loop 1. The LLM is used to run structured reconciliations on financials, flag discrepancies, draft the deliverables in scope (memos, schedules, draft returns, advisory notes), and prepare responses to client questions. Iteration is expected as clients respond and findings evolve. Move on once draft deliverables and responses are technically sound and ready for client review.

Prompt · Financials Review & Discrepancy Flagging
Context
The User has received client financials and supporting documents and needs a structured review for completeness, internal consistency, and reasonableness before drafting deliverables. This prompt produces a findings log and clarifying questions list. User Provides
- Loop 1 artefacts (account profile, intake matrix): [LOOP 1 ARTEFACTS]
- Financials summary (TB, P&L, BS, key roll-forwards): [FINANCIALS]
- Supporting items (payroll summary, debt schedule, fixed assets, etc.): [SUPPORTING]
- Prior-year comparatives: [PRIOR-YEAR DATA] Ask
Perform reconciliations and reasonableness checks (intercompany ties, TB-to-financials, P&L variance vs. prior year, balance sheet roll-forwards, payroll-to-GL). Flag discrepancies, anomalies, and items requiring client clarification. Return Format
1. Reconciliation Results (table: Check | Expected | Actual | Variance | Status)
2. Variance & Anomaly Findings (with hypothesis and recommended action)
3. Clarification Questions for Client (grouped, numbered)
4. Items Ready to Advance to Drafting Constraints / Quality Bar
- Cite the specific schedule or line for each finding.
- Do not assume causes for variances; suggest hypotheses to verify.
Prompt · Draft Deliverable Builder
Context
The User is preparing the in-scope client deliverable(s) (e.g., tax return, advisory memo, planning summary, response to notice). This prompt generates a structured draft grounded in validated data, with assumptions and workpaper references. User Provides
- Validated financials and resolved discrepancies: [VALIDATED INPUTS]
- Deliverable type and audience: [DELIVERABLE TYPE]
- Applicable tax/accounting treatments to apply: [TREATMENTS]
- Client-specific positions or preferences from prior years: [HISTORICAL POSITIONS] Ask
Draft the deliverable with clear structure, applied treatments, key computations, assumptions, and workpaper references. Highlight areas requiring User judgement before client release. Return Format
1. Draft Deliverable (structured per type)
2. Key Computations & Assumptions
3. Workpaper Reference Index
4. Reviewer Watch-Items (areas needing User judgement)
5. Open Questions to Resolve Before Client Review Constraints / Quality Bar
- Cite authority for any tax treatment applied.
- Keep assumptions visible and easy to amend.
Prompt · Client Question & Revision Response Pack
Context
The User has shared a draft deliverable with the client and received questions or revision requests. This prompt structures responses, identifies which changes are accepted, and documents impacts on the deliverable. User Provides
- Current draft deliverable: [CURRENT DRAFT]
- Client questions and revision requests (sanitised): [CLIENT FEEDBACK]
- Constraints (technical, regulatory, fee/scope): [CONSTRAINTS] Ask
For each client item, propose a response, indicate whether it should be accepted/declined/clarified, describe the resulting change to the deliverable, and flag any items that require User judgement or escalation. Return Format
1. Response Matrix (table: Client Item | Proposed Response | Disposition | Deliverable Impact | Judgement Required?)
2. Updated Sections of Deliverable (drafted)
3. Items to Escalate or Discuss Live with Client Constraints / Quality Bar
- Maintain professional, neutral tone.
- Do not concede technical positions without authority-backed rationale.
Review before proceeding
  • Reconciliation findings have been resolved or appropriately escalated
  • Draft deliverables are technically accurate and consistent with prior positions
  • Reviewer watch-items have been dispositioned
  • Client questions have proposed responses with clear dispositions
  • Draft is ready for client review or final sign-off
Outputs
  • Reconciliation findings and clarification log
  • Draft deliverable(s) with workpaper references
  • Client question response matrix
  • Updated deliverable sections
Loop 3: Synthesize & package

Finalize, file, and close out the deliverable with clear client communication, documented approvals, and archived workpapers.

The User brings the near-final deliverable, client revisions, and reconciled workpapers from Loop 2. The LLM is used to craft the final client communication, the filing/submission checklist, and the archival index. Iteration is expected on tone and emphasis. Move on once the User has signed off on all closeout artefacts.

Prompt · Finalization, Filing & Closeout Package
Context
The User is ready to finalize the deliverable, secure approvals, file/submit as required, and archive workpapers. This prompt produces the closeout package, client communication, filing checklist, and archive index, without introducing new technical positions. User Provides
- Final deliverable and resolved client feedback: [FINAL DELIVERABLE]
- Filing/submission requirements (channels, authorities, signatures): [FILING REQS]
- Client communication preferences and tone: [TONE / PREFERENCES]
- Workpapers and supporting documents to archive: [WORKPAPER INDEX] Ask
Produce (1) a final client delivery message and summary, (2) a filing/submission checklist with owners and confirmations, (3) an archive index of workpapers, and (4) a forward-looking follow-up list for the next engagement cycle. Return Format
1. Final Client Message (concise, professional, action-oriented)
2. Executive Summary of Deliverable (key results, decisions, next steps)
3. Filing & Submission Checklist (table: Item | Owner | Status | Confirmation)
4. Workpaper Archive Index (table: Document | Location | Retention Note)
5. Forward Follow-Up List (planning items, next deadlines, account growth opportunities) Constraints / Quality Bar
- Match the client's preferred tone.
- Do not introduce new positions or analysis not already in the deliverable.
Review before proceeding
  • Final client message accurately reflects the deliverable and next steps
  • All required signatures and approvals are obtained
  • Filings/submissions are confirmed with proof retained
  • Workpaper archive index is complete
  • Forward follow-up list captures planning and account opportunities
Outputs
  • Final client delivery message and executive summary
  • Filing and submission checklist with confirmations
  • Workpaper archive index
  • Forward follow-up list

Develop tax planning strategies

Loop 1: Frame & shape inputs

Define planning objectives, constraints, and the input baseline needed to evaluate strategy options for the client.

The User typically brings a client situation (entity structure, recent financials, prior returns, contemplated transactions) and a planning question or savings goal. The LLM is used to translate this into a planning brief, identify which inputs and authorities are needed, and surface assumptions and constraints. Iteration is expected as the User refines objectives or receives more client information. Move on once the planning brief, baseline facts, and input checklist are confirmed.

Prompt · Planning Brief & Baseline Framing
Context
The User is a Tax Manager developing planning strategies for a client. Before research and modeling begin, a structured planning brief is needed that captures objectives, constraints, current tax posture, and the scope of strategies to consider. This prompt exists to align the work and prevent premature jumping to recommendations. User Provides
- Client entity structure and ownership summary: [STRUCTURE]
- Current financial and tax posture (recent results, ETR, key positions): [POSTURE]
- Planning objectives and any savings target or constraint: [OBJECTIVES]
- Contemplated transactions or life events (M&A, succession, exits, relocations): [EVENTS]
- Risk appetite and reputational considerations: [RISK APPETITE]
- Timeline and decision deadlines: [TIMELINE] Ask
Build a planning brief that (1) restates objectives in measurable terms, (2) maps the current tax baseline, (3) identifies candidate strategy themes worth researching, (4) lists assumptions and constraints, and (5) defines success criteria for any recommendation. Return Format
1. Objectives (measurable, prioritized)
2. Current Tax Baseline (entities, ETR, key positions, carryovers)
3. Candidate Strategy Themes (with brief rationale, not yet researched in depth)
4. Assumptions & Constraints
5. Success Criteria & Decision Inputs
6. Open Questions Before Research Constraints / Quality Bar
- Do not commit to specific strategies yet; only theme-level candidates.
- Distinguish facts from assumptions explicitly.
Prompt · Input & Research Checklist
Context
The User has a planning brief and now needs a precise list of inputs and authorities required to research and model the candidate strategy themes. This prompt produces an input checklist and research scoping plan. User Provides
- Validated planning brief from prior prompt: [PLANNING BRIEF]
- Documents and data already available (summaries): [AVAILABLE INPUTS]
- Jurisdictions in scope: [JURISDICTIONS] Ask
For each candidate strategy theme, list (a) the client inputs required, (b) the authorities and guidance to consult, and (c) the modeling variables needed. Identify what is missing and propose how to obtain it. Return Format
1. Per-Theme Input Checklist (table: Theme | Client Inputs | Authorities | Modeling Variables | Status)
2. Cross-Theme Common Inputs
3. Missing Items & How to Obtain Them
4. Research Scoping Plan (sequence, depth, time estimate) Constraints / Quality Bar
- Keep checklists specific; avoid generic 'tax law' references.
- Flag any theme that is not viable to research due to missing critical inputs.
Review before proceeding
  • Objectives and success criteria are measurable and client-aligned
  • Candidate strategy themes are reasonable given the facts
  • Assumptions and constraints are explicit
  • Input checklist and research scope are achievable in the timeline
  • Critical missing inputs have a plan to be obtained
Outputs
  • Planning brief
  • Per-theme input and research checklist
  • Assumptions and constraints log
Loop 2: Execute heavy lifting

Research authorities, model scenarios, and produce a comparative evaluation of viable strategies with quantified impacts and risks.

The User brings the validated planning brief and inputs from Loop 1. The LLM is used to perform targeted research on each theme, build scenario models with assumptions exposed, and compare strategies on savings, cashflow, risk, and implementation complexity. Iteration is expected as assumptions change, new client data arrives, or authorities shift the conclusion. Move on once a short list of viable strategies has been quantitatively and qualitatively evaluated.

Prompt · Targeted Research & Authority Map
Context
The User needs grounded research on each candidate strategy theme, with authorities cited and current guidance reflected. This prompt produces a research memo per theme that supports or rules out the theme. User Provides
- Planning brief and input checklist from Loop 1: [LOOP 1 ARTEFACTS]
- Themes to research (selected from candidates): [THEMES]
- Jurisdictions and relevant client positions: [CONTEXT] Ask
For each theme, produce a research memo covering applicable Code sections, regulations, rulings, and recent guidance; eligibility requirements; documentation requirements; common pitfalls; and a preliminary go/no-go view with rationale. Return Format
For each theme:
1. Theme & Hypothesis
2. Authorities & Guidance (with citations)
3. Eligibility & Requirements
4. Documentation Requirements
5. Common Pitfalls & Audit Risk Areas
6. Preliminary Go / No-Go / Needs-More-Info with rationale Constraints / Quality Bar
- Cite specific authority; flag when guidance is unclear or unsettled.
- Note where state/local rules diverge from federal.
Prompt · Scenario Modeling & Comparative Evaluation
Context
The User has research memos and needs quantitative scenario modeling for the viable strategies. This prompt produces structured scenario models and a comparative evaluation across savings, cashflow, complexity, and risk. User Provides
- Research memos from prior prompt: [RESEARCH MEMOS]
- Baseline financial and tax projections: [BASELINE]
- Strategy-specific modeling assumptions (rates, timing, elections): [ASSUMPTIONS]
- Time horizon for evaluation: [HORIZON] Ask
For each viable strategy, build a scenario model showing baseline vs. strategy outcomes (tax, cashflow, ETR, basis, carryovers) over the horizon. Then produce a comparative table evaluating strategies on savings, cashflow timing, implementation complexity, compliance burden, audit risk, and reversibility. Return Format
1. Per-Strategy Scenario Model (table: Year | Baseline | Strategy | Delta, for tax, cashflow, ETR)
2. Key Drivers & Sensitivities
3. Comparative Evaluation Matrix (Strategy | Savings | Cashflow Timing | Complexity | Compliance Burden | Audit Risk | Reversibility)
4. Recommended Short List with rationale
5. Assumptions Used (visible and editable) Constraints / Quality Bar
- Make assumptions explicit and editable.
- Show sensitivities for assumptions that materially move outcomes.
- Do not present results without uncertainty acknowledgement.
Review before proceeding
  • Authorities cited are accurate and current
  • Modeling assumptions are realistic and align with client facts
  • Sensitivities to key assumptions have been tested
  • Short-listed strategies are compliant and feasible
  • Risks and reversibility are appropriately reflected
Outputs
  • Per-theme research memos
  • Scenario models with assumptions
  • Comparative evaluation matrix
  • Short-listed strategy set
Loop 3: Synthesize & package

Package the recommended strategies into a client-ready advisory deliverable with implementation roadmap and monitoring plan.

The User brings the short-listed strategies, models, and evaluation matrix from Loop 2. The LLM is used to translate these into a prioritized recommendation memo, implementation roadmap, and post-implementation monitoring plan. Iteration is expected on tone, level of technical detail, and emphasis. Move on once the User is ready to present the recommendations to the client.

Prompt · Recommendation Memo & Implementation Roadmap
Context
The User has short-listed strategies, models, and a comparative evaluation. This prompt produces a client-facing recommendation memo with an implementation roadmap and a monitoring plan, calibrated to the client's risk appetite and decision style. User Provides
- Loop 2 artefacts (research, models, comparative matrix): [LOOP 2 ARTEFACTS]
- Client decision style and risk appetite: [CLIENT PROFILE]
- Internal and external resources available for implementation: [RESOURCES]
- Required documentation and reporting requirements: [DOC REQS] Ask
Produce (1) a prioritized recommendation memo with rationale, quantified impact, and risks, (2) an implementation roadmap with phases, owners, milestones, and dependencies, (3) a documentation checklist, and (4) a post-implementation monitoring plan with review triggers. Return Format
1. Executive Summary (recommendations and expected impact)
2. Prioritized Recommendations (table: Strategy | Priority | Estimated Savings | Key Risks | Decision Needed)
3. Implementation Roadmap (phases, owners, milestones, dependencies)
4. Documentation & Compliance Checklist
5. Monitoring Plan (KPIs, review cadence, triggers for revisiting strategy)
6. Open Questions / Decisions Required from Client Constraints / Quality Bar
- Keep the executive summary concise; detail goes into appendices.
- Tie every recommendation to authority and to the model output.
- Be explicit about decisions the client must make and by when.
Review before proceeding
  • Recommendations align with client objectives and risk appetite
  • Quantified impacts trace back to the validated models
  • Implementation roadmap is realistic with clear owners
  • Documentation and monitoring plans are complete
  • Client decisions and timing are explicit
Outputs
  • Recommendation memo
  • Implementation roadmap
  • Documentation and compliance checklist
  • Post-implementation monitoring plan
What to do next

Where accounting leaders should focus

Four moves follow from the evidence.

01

Treat it as function-wide redesign

No role is untouched and none is fully exposed. The response is to redesign how the function works end to end, rather than ring-fence a few roles.

02

Start with the processing core

Payables, payroll, and reconciliation carry the highest automation potential and the most routine time today. They are the fastest place to prove value at low risk.

03

Reinvest freed time into judgement

The skills that hold the function together are relational. Recovered capacity should go there by design. Plan the reinvestment rather than letting it default.

04

Adopt through workflows, not tools

Exposure is led by tool use and reasoning over connected data. Value comes from designed and reviewed workflows like those above, not from a model used on its own.

Method & sourcing

About this analysis

The twelve roles were selected as a representative cross-section of the accounting function and analysed from job-description evidence. Each role was decomposed into tasks and actions, and every action assessed against four AI capabilities to produce task-level and role-level automation potential. Use-case workflows were generated for every role and checked for the three-loop collaboration pattern.

Automation potential describes the share of measured task time open to AI support, with a person remaining the lead. It points to opportunity for redesign, not a forecast of headcount. Figures are point-in-time and will move as the work and the tools evolve.

Tempris builds targeted action that grows the business while future-proofing the workforce.