Skip to content

AI-assisted internal-audit fieldwork

From audit evidence to answers you can review.

AI-assisted internal-audit fieldwork for Indian enterprises. Explore a worked example of a procedure, its sources and the decisions that stay with your team.

For internal-audit teams at Indian enterprises.

Interactive example · synthetic data
DEMO-0042Fail

Amount mismatch

₹5,000

above the approved purchase order

Invoice subtotal
₹1,25,000
Approved PO
₹1,20,000
Inspect the source
The rule behind the result

Synthetic Procurement Policy v3 · §4.2
Tolerance ₹1 · Review incomplete

Representative interface · P2P amount check v3.

01 / The evidence, in focus

Follow one invoice
from source to review.

Select a step. Inspect the evidence. Keep the conclusion in human hands.

Interactive example · synthetic data. Representative interface, not a product capture.

Interactive example · synthetic data
P2P — Q1 FY27DEMO-RUN-018 · Recipe v3

Result / DEMO-0042

Fail

The amounts don’t match.

Amount mismatch. The invoice subtotal exceeds the approved purchase order by ₹5,000; tolerance ₹1.

Expected · approved PO
₹1,20,000
Actual · invoice subtotal
₹1,25,000
Difference
₹5,000

Awaiting reviewer confirmation. Selecting an example does not approve a result.

Source TraceVersion 1 · excerpt

Illustrative invoice

Aranya Office Supplies

DEMO-0042.pdf · Page 1
Invoice date: 15 April 2026 · PO-0042


Office equipment₹1,25,000
Subtotal₹1,25,000

Trace: page 1 · subtotal region
Original: “1,25,000.00” → normalised: ₹1,25,000
Method: illustrative extraction · confidence: 99%
Corrections: none in this fixture

Rule: Synthetic Procurement Policy v3 · §4.2
P2P amount check · v3 · Allowed tolerance ₹1

Representative interface. Invented evidence and policy; no live audit execution.
Read all six steps and source references
  1. Documents

    The invoice is primary evidence; the approved purchase order is the amount authority. An ERP payment export defines the population. Goods receipts support the procedure, reference data supplies mappings, and the synthetic procurement policy defines the criteria.

  2. Tests

    The Audit Test Recipe holds Extract, Reconcile and Checks. Match on the PO reference; compare invoice subtotal with the approved PO amount using a ₹1 tolerance and Synthetic Procurement Policy v3 §4.2. Recipe v3 is approved only within this illustration; conclusions still require human review.

  3. Runs

    DEMO-RUN-018 uses Recipe v3 and input versions v1. Of 200 received records, 10 are excluded and 190 eligible. Execution completes for 180; 6 await evidence and 4 have processing errors. In the product model, completed Run history is immutable; new versions do not rewrite earlier work.

  4. Review

    DEMO-0042: ₹1,25,000 invoice subtotal versus ₹1,20,000 approved PO. The ₹5,000 difference exceeds ₹1 tolerance: Fail / Amount mismatch. Trace the invoice to page 1 and the PO to Orders!H43. Reviewer confirmation remains outstanding.

  5. Findings

    A finding groups related confirmed exceptions into an audit issue with criteria, condition, cause, impact, recommendation and sign-offs. DEMO-0042 remains unconfirmed: no finding has been raised and there are zero related confirmed exceptions in this illustration.

  6. Working Papers

    The working paper brings scope, population, procedure, exceptions, source references, comments and sign-offs together. DEMO-RUN-018 remains Draft / review incomplete. Moving through this example does not approve a result or create an export.

Source Traces for DEMO-0042

Primary evidence: DEMO-0042.pdf · Page 1 · subtotal region · v1. Original “1,25,000.00” → ₹1,25,000. Illustrative extraction; none in illustration.

Source of truth: purchase-orders.xlsx · Orders!H43 · v1. Original “120000” → ₹1,20,000. Illustrative cell read; none in illustration.

Synthetic Procurement Policy v3 · §4.2 · effective 1 April 2026. Compare invoice subtotal with approved PO amount. Absolute difference must not exceed ₹1. If the approved PO is unavailable, report Insufficient evidence. Invented company policy, not a regulatory requirement.

Missing evidence is reported separately unless the configured procedure explicitly defines its absence as a failure.

Copilot drafts and explains. Auditors approve Recipes and conclusions; overrides require a reason and an attributable record.

02 / The procedure behind the result

Your audit procedure,
made repeatable.

Define the objective, population, sources, comparisons and approval requirements before interpreting a result. Extract, Reconcile and Checks belong inside Tests.

Explore Audit Test Recipes

Illustrative procedure → draft Recipe

“Compare the invoice subtotal with the approved purchase-order amount. Investigate differences above ₹1.”
Scope
Objective: establish amount agreement.
Risk: invoice overstatement.
PO-backed invoices · Q1 FY27 · invoice, PO and supporting sources.
Logic
Extract subtotal and PO reference. Reconcile to the approved PO as amount authority. Check absolute difference against ₹1; retain all six result states and source Traces.
Control
Set exception severity, reviewer requirements and output before approval. Record author, approver, version and effective date.

Synthetic Procurement Policy v3 · §4.2 · effective 1 April 2026

P2P amount check · v3 · Approved in illustration

DEMO-RUN-018 · Awaiting reviewer confirmation

A newly interpreted procedure is a draft until a person approves it. This illustration’s existing v3 approval is separate from approval of its results.

03 / Coverage and judgement

Know what was tested.
See what needs attention.

One configured check per eligible record. These are illustrative population counts, not an assurance or accuracy measure.

180 / 190

eligible records completed execution.
Completion does not resolve human review.

Received
200
Excluded
10
Eligible
190
Completed
180
Awaiting evidence
6
Processing errors
4
  • Pass157
  • Fail17
  • Insufficient evidence6
  • Needs human review6
  • Not applicable0
  • Processing error4

Six states across 190 eligible records. Excluded records are separate from Not applicable outcomes. Missing evidence remains separate unless the configured completeness procedure makes its absence a failure.

AI assists. Code compares. People approve.

AI extracts and drafts; deterministic code performs the configured arithmetic. A reviewer assesses sources, resolves uncertainty and owns the conclusion.

Explore review and findings
Inspect a proposed reviewer decision

Illustrative review note · not submitted

“The difference needs investigation. Retain the amount mismatch until supporting evidence establishes a reason to change it.”

Preparer
Proposes a correction or override with its reason and cited evidence.
Independent reviewer
Assesses the proposal under the maker-checker requirement. A preparer cannot provide their own independent sign-off.
This record
DEMO-0042 · Fail / Amount mismatch. Awaiting reviewer confirmation. No override has been applied.

A decision needs an attributable activity record. Opening this note does not record approval.

04 / The work your team performs

Built around the procedures
your team repeats.

Follow P2P amount comparisons, sanction versus disbursement checks, or independent review of journal-entry controls. Each example uses its own supplied evidence and criteria. The examples below are invented illustrations, not current product Runs.

01 / DEMO-0042

Procure-to-pay

Invoice ↔ approved PO ↔ goods receipt ↔ ERP export

Configured check
Compare invoice subtotal with the approved PO amount.
Possible exception
₹5,000 above the approved amount; reviewer confirmation outstanding.
Review output
Record-level exception with source Traces and rule citation. Draft / review incomplete.
Inspect this example’s sources

DEMO-RUN-018 · Synthetic Procurement Policy v3 · §4.2

DEMO-0042.pdf · Page 1 · subtotal region
Primary evidence · ₹1,25,000

purchase-orders.xlsx · Orders!H43
Source of truth · ₹1,20,000

Explore procure-to-pay

02 / DEMO-LN-0011

Credit and loan files

Sanction ↔ disbursement ↔ supplied borrower evidence

Configured check
Compare disbursement with the sanctioned amount.
Possible exception
₹36,00,000 disbursed against ₹35,00,000 sanctioned; a ₹1,00,000 difference.
Review output
Record-level exception with source Traces and rule citation. Draft / review incomplete.
Inspect this example’s sources

DEMO-CREDIT-001 · Synthetic Lending Policy v1 · §2

DEMO-LN-0011.pdf · Page 1 · sanctioned amount
Source of truth · ₹35,00,000

demo-disbursements.csv · Row 12 · disbursed_amount
Primary evidence · ₹36,00,000

Explore credit and loan files

03 / DEMO-JEA-002

Financial and IFC/SOX controls

Journal-entry control record ↔ sign-off evidence ↔ control procedure

Configured check
Check that the control reviewer differs from the performer.
Possible exception
Both recorded as DEMO-PERSON-01; independent review is not evidenced.
Review output
Record-level exception with source Traces and rule citation. Draft / review incomplete.
Inspect this example’s sources

DEMO-IFC-001 · Synthetic Control Procedure v1 · §2

DEMO-JEA-002.pdf · Page 1 · performed by
Primary evidence · DEMO-PERSON-01

DEMO-JEA-002.pdf · Page 1 · reviewed by
Primary evidence · DEMO-PERSON-01

Explore financial and ifc/sox controls

Operational and compliance procedures follow the same supplied-evidence model. Scope, applicable sources and suitability need evaluation; document checks do not replace observation or professional judgement.

05 / What the reviewer receives

The working paper carries
the evidence with it.

A working paper brings scope, procedure, coverage, results, exceptions, evidence references and sign-offs together. The illustrated P2P work remains Draft / review incomplete; no genuine download is available in this example.

Select a contents annotation to inspect the reasoning and remaining review work.

Explore working papers

Illustrative contents · not a product export

P2P — Q1 FY27

Draft / review incomplete

DEMO-RUN-018 · DEMO-0042

01Scope and population

P2P — Q1 FY27 · 1 April–30 June 2026. 200 received, 10 excluded, 190 eligible; 180 completed, 6 awaiting evidence, 4 processing errors.

02Procedure and criteria

P2P amount check · v3. Synthetic Procurement Policy v3 · §4.2. Compare invoice subtotal with approved PO amount; tolerance ₹1.

03Exception register

DEMO-0042 · Amount mismatch. Expected ₹1,20,000; actual ₹1,25,000; difference ₹5,000. Confirmation outstanding.

04Source references

DEMO-0042.pdf · page 1 · subtotal region · v1. purchase-orders.xlsx · Orders!H43 · v1. Original and normalised values stay with their Traces.

05Comments and sign-offs

Review incomplete. No confirmed finding, submitted override or final sign-off. Approved working papers lock in the product model; later changes create a new version. This illustration is a draft.

This HTML explanation contains invented data. Discuss the required output in a demo; it is not a downloadable working paper.

06 / Governance and supporting proof

Built for evidence
your team is accountable for.

Evaluate the controls alongside your procedure. These questions describe what to verify for the product environment you will use.

Explore security evaluation
Access and approval
Verify engagement permissions, independent review and reasoned overrides with your team’s roles.
Version history and activity
Inspect retained input and Recipe versions, immutable completed Run history and attributable decisions. Integrity of a snapshot does not mean the audit passed.
Data handling and deployment
Confirm hosting regions, provider data flows, retention and contractual commitments for the environment being evaluated.

Keep policies current. Keep answers tied to sources.

Knowledge Hub holds versioned sources; Copilot’s role is to draft and explain with citations. People approve Recipes and conclusions.

Explore Knowledge Hub and Copilot

07 / Start with one procedure

See your audit workflow
in DocRack.

Request a demo to discuss your evidence, configured rules and reviewer requirements. Submitting the form requests a conversation; it does not reserve a calendar slot.

Describe the procedure. Please keep confidential audit evidence out of the enquiry form.