Main site

Accounting AI agent · Disclosure drafting

Disclosure-Drafting & Reporting Assistant (Notes to the Accounts)

Draft the notes from approved figures and policies, cross-cast every figure to the statements — held for management approval, with evaluation and guardrails built in.

Typical delivery
4–6 weeks
Deployment
Your stack
Human review
Before approval
After launch
Agent Care

What this agent does

Drafts the notes and ties them back

In

  1. 01

    Ingest the trial balance, the primary statements and last year's approved accounts.

  2. 02

    Normalise each note into entity, period, framework and the figures it has to carry.

Reason

  1. 03

    Draft the note from the approved policy, the agreed template and the figures behind it.

  2. 04

    Roll prior-year comparatives forward and re-check them against any restatement.

  3. 05

    Work the disclosure checklist for that framework, period and entity.

Decide

  1. 06

    Flag notes that no longer agree to the statement, the ledger or each other.

  2. 07

    Route materiality, judgement and new-disclosure questions to management.

Out

  1. 08

    Retain each figure, its source, the drafted text, the checklist result and the decision.

  2. 09

    Write into the reporting file only inside the approval boundaries agreed during implementation.

Product statement

The agent drafts and cross-checks inside the approval boundaries agreed during implementation; it judges no materiality question, and management approves the financial statements.

Example workflow

One note, end to end

  1. 1Note openedThe reporting calendar, the disclosure checklist or a movement in the trial balance
  2. 2Scope resolvedEntity, period, framework, the thresholds management set and last year's approved wording
  3. 3Note draftedFigures taken from the ledger, comparatives rolled forward, text built from the approved policy and template
  4. 4Controls appliedCross-cast, checklist coverage, comparative and cross-reference checks and confidence threshold

No human action required

Stages 1 to 4 run without a person in the loop — nothing has entered the reporting file yet, so the lane stays empty until the gate.

5DecisionSplits on the confidence and note-class gate
High confidence

Joins the reporting file as a cross-cast draft.

Low confidence

Held with the figure it could not agree.

Note owner review

The note is held with its figures, their source, the checklist result and the break that stopped it.

Approve · Amend · Refer up

Approved — handed back

  1. 6Note enters the reporting fileApproved before the statements are issued; the agent settles nothing about what has to be disclosed
  2. 7Outcome evaluatedCross-cast accuracy, checklist coverage, text rewritten in review, tagging validation and business outcome

Rewrites

Text rewritten and referred up in review is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Whether an item is material enough to disclose.
    • Anything issued to shareholders or a regulator.
    • New or amended accounting-policy wording.
    • Going-concern wording and material uncertainties.

  • Automation boundary

    Agent acts unaided

    • ✓Draft note text from the approved policy and template.
    • ✓Cross-cast every figure to the statements and the ledger.
    • ✓Roll comparatives forward and check them for restatement.
    • ✓Work the checklist and refer open questions up.

    Text and figures reach the reporting file only inside the approval boundaries agreed during implementation.

    • Judgements, estimates and the sensitivities shown.
    • Related-party relationships and subsequent events.
    • Restating a comparative or a prior-period error.
  • Changes to templates, checklists or thresholds.

Example output

One note, annotated

Everything the agent drafts is attached to the figure and the policy it came from.

Drafted note · single disclosure

Illustrative example
Note
Trade and other payables
Entity
Reporting group
Period
FY25
Cross-cast
Agrees to the balance sheet
Confidence
92%
Held for
A restated comparative

As received

Exactly as the reporting calendar sets them — the agent does not decide a note is needed.

Evidence used

  • Year-end ledger balance
  • Last year's signed note
  • Approved policy wording

Why it stops here

The current figure agrees; the comparative does not match last year's approved note.

What the score decides

Below the configured threshold the note is held for its owner instead of entering the file.

Action

ApproveAmendRefer up

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

All open notes

From the reporting calendar

03Drafting

Apply your own policy wording

Use the approved accounting policies, the group's templates and the note wording signed off last period.

01Approved path

Figures agreed before review

Each amount in a drafted note is cast, cross-cast to the primary statements and traced back to the ledger.

02Human review

Name the gaps the draft cannot close

Disclosure points the draft cannot answer, and wording that moved since last period, are listed for the reviewer.

04

Build an evidence trail

Retain each figure, its source, the drafted text, confidence, evaluator result and the approval decision — on both paths.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Disclosure management

    • Workiva
    • CCH Tagetik
    • Oracle Narrative Reporting
    • OneStream

  • ERP, GL & consolidation

    • SAP
    • Oracle
    • NetSuite
    • Sage Intacct
    • trial-balance exports

  • Checklists & workpapers

    • Caseware
    • CCH Accounts Production
    • Checklist libraries
    • workpaper exports

Agent

Disclosure-drafting assistant

  • Reads the figures
  • Drafts and cross-casts
  • Routes to approval

  • Documents & tagging

    • iManage
    • SharePoint
    • Tagging tools
    • iXBRL/ESEF validators

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the accounts

Each control wraps the one inside it. A note clears every layer before it can enter the reporting file.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeRestrict automation if evaluations or production signals degrade.Roll back
L5Version monitoringTrack ledger re-extracts, standard updates, model and prompt changes.Track
L4TraceabilityRecord figure, source, draft, checklist result, evaluation and rewrite.Record
L3Management approvalDefine which notes may enter without management approving.Gate
L2Source-bound textWording must trace to a figure, a policy or an approved template.Restrict
L1Cross-cast checksEvery figure in a note must agree to the statement and the ledger.Cross-cast
Model coreNote drafted — figures, text, comparatives, checklist result and confidence
L1 – L2
Decide whether the drafted note may stand
L3
Decides whether management approves it
L4 – L5
Keep the record and the figures current
L6
Pulls automation back when signals degrade

How Nestack evaluates it

Evaluate the full workflow — not only the finished note.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the note the reviewer reads

  1. E1Final-output evaluationDid the note carry what the framework requires?
  2. E2Cross-cast evaluationDoes every figure still agree to the statement?
  3. E3Coverage evaluationWas every requirement covered, and does the tagging validate?
  4. E4Comparative evaluationDo prior-year figures match last year's accounts?
  5. E5Slice evaluationHow does performance change across specific note cohorts?
  6. E6Business outcomeHow much drafted text survived review unchanged?

Floor — the outcome the client pays for

Failure modes

Where each failure originates in the agent

Seven failure modes plotted against the five stages of the agent lifecycle.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    2 modes

    • DR-01

      Comparative unchecked

      Last year's figures roll forward despite a restatement.

    • DR-02

      Policy not on file

      No source states the policy the entity applies now.

    Stage gathers

    Trial balance, statements, policies and last year's notes

  2. 02 · Reasoning

    2 modes

    • DR-03

      Boilerplate over entity facts

      Template wording outlives the policy, or the entity's facts.

    • DR-04

      Requirement missed

      A standard's new disclosures are missed or wrongly dated.

    Stage proposes

    The drafted note, its figures, comparatives and confidence

  3. 03 · Tool / write

    1 mode

    • DR-05

      Written before approval

      Text enters the reporting file before management approves.

    Stage writes

    Only where file access and approval boundaries allow it

  4. 04 · Output

    1 mode

    • DR-06

      Reference or tag misdirected

      A renumbered note or mis-tagged figure points elsewhere.

    Stage returns

    The note the reviewer and the reporting file see

  5. 05 · Change / Version

    1 mode

    • DR-07

      Note stops agreeing

      A late adjustment moves the statement under the note.

    Stage tracks

    Ledger re-extracts, standard updates and prompt changes

  • Sev-1 · unapproved wording could be issued
  • Sev-2 · wrong figure or stale wording in a note
  • Sev-3 · drafting degrades, note held back
See the complete Accounting failure-mode catalogue

Affected slices

The notes that change are the notes that fail

Rolled-forward notes behave; the ones written for the first time, or against a restated comparative, carry most of the cross-cast and comparative errors. Nestack reports performance by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
First-time adoption notes6.4%3.7×Review
Restated comparatives4.4%2.6×Review
Judgement and estimate notes3.1%1.8×Watch
Rolled-forward routine notes1.1%0.6×Normal

Bar: lift vs. routine-note baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

Every rewrite ends up in the template

A rewrite in review does not stop at this period's note — the wording, the check or the checklist entry behind it is changed too.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Rewrites, cross-cast breaks or checklist gaps rise in a note cohort.

  2. 02

    Diagnose

    Failure traced to the figure source, the template, the checklist or the comparative.

  3. 03

    Improve

    Template wording or a checklist entry is corrected — approved and version-linked.

  4. 04

    Verify

    Affected notes are re-drafted against the current statements.

  5. 05

    Learn

    The approved wording becomes the template; the break becomes a standing check.

  6. Learn → Detect

    The return edge. Next period starts from corrected wording and one more cross-cast check.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, figures and templates, drafting workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Workflow discovery and approval-boundary definition.Week 1
  2. 02Source-system and platform assessment.Week 1, Week 2
  3. 03Note inventory, framework and threshold mapping.Week 1, Week 2
  4. 04Figure extraction and note templates.Week 2
  5. 05Drafting, cross-cast and comparative checks.Week 2, Week 3
  6. 06Confidence scoring and referral routing.Week 3
  7. 07Management approval workflow.Week 3, Week 4
  8. 08Reporting-platform and tagging integration.Week 5
  9. 09Evaluation suite and regression cases.Week 4, Week 5
  10. 10Checklist coverage and drafting guardrails.Week 4
  11. 11Observability and trace instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them. No tier lets the agent decide what must be disclosed.

Capability✓ in scope · — not at this tierPilotOne set of notesProductionProduction integrationAdvancedGroup and multi-framework
Introduced at Pilot
Drafts traced to approved sources✓✓✓
Cross-cast to the statements✓✓✓
Management approval✓✓✓
Baseline evaluation✓✓✓
Introduced at Production
Disclosure-checklist coverage—✓✓
Comparative and restatement checks—✓✓
Approved write actions—✓✓
Observability and evaluation—✓✓
Introduced at Advanced
Group and multi-framework notes——✓
iXBRL and ESEF tagging support——✓
Enterprise controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on the notes in scope, reporting frameworks, source systems, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01The notes in scope and the framework they follow

    Note inventory, framework and threshold mapping

    Week 1

  2. 02Who approves each note and who approves the accounts

    Approval-boundary definition and write controls

    Week 1

  3. 03Access to the ledger, consolidation and reporting file

    Source-system and reporting-platform assessment

    Week 2

  4. 04Last year's approved accounts and your note templates

    Note templates and comparative roll-forward checks

    Week 2

  5. 05Your checklist and the thresholds management set

    Checklist coverage, cross-cast checks and referral routing

    Week 3

  6. 06The notes that were rewritten most last period

    Evaluation suite, regression cases and failure-mode testing

    Week 4

  7. 07Named reviewers or test users

    Management approval workflow, then a pilot note set and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Phases are drawn over the weeks they actually occupy. Drafting waits on the note inventory agreed in week 1.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Workflow discovery, approval boundary and note inventory

W2Figure access, note templates and last year's wording

W3Drafting workflow, cross-cast checks and approval controls

W4Evaluation suite, checklist coverage and failure modes

W5Reporting-platform integration, pilot notes and corrections

W6A live note set drafted and approved, then Agent Care handover

Reading the band
Bars cover only the weeks their work is named in. Evaluate and Pilot share week 5: the first live notes are drafted while the suite still runs.
At the end of W6
A live set of notes has been drafted, cross-cast and approved, then Agent Care watches the cross-casts, the checklist and the rewrite rate.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Accounting AI agent

Build a disclosure-drafting assistant around your reporting cycle.

Show us your note set, last year's approved accounts and your disclosure checklist, and tell us who approves what. We'll draft one note against your own checklist and agree what the agent may never settle.

Typical delivery
4–6 weeks
Deployment
Your reporting stack
Human review
Before management approves
After launch
Agent Care available

Nestack Agents · Disclosure-drafting & reporting assistant

AGT-ACC-13 · Agent Care available after launch

Build a disclosure-drafting assistant

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your books

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.