traack

AI agents for internal auditing and SOX compliance teams

Traack agents do the busy work of control testing so your team spends its time on judgement calls. Upload your RCM and evidence, and the agent runs continuous full-population testing, drafts the workpaper and flags every exception. SOC 2 compliant, zero data retention.

Q3 manual journal entry approval testing
XLSRCM_FY26.xlsx
Test control JE-03 for Q3: every manual journal entry approved by someone other than the preparer before posting, dual approval above $10,000, and support attached. Population is the NetSuite manual JE export; approvals are in the NetSuite workflow log. Sample 25 for evidence and list every exception.

JE-03 in the RCM is the manual journal entry review control, tested quarterly with a 25-item sample. I'll screen the full Q3 population first, then pull evidence for the sample and every entry the screen flags.

Pass — all four attributes1,273$47.1M
Exception — one or more attributes failed11$1.1M
Sample — evidence pulled and matched25$0.9M
entriesamount
XLS212 KB · 5 sheets·

Eleven entries need your judgement. Four were approved by their own preparer (all by the same user, R. Meyer, in August). Seven posted before the approval timestamp; six of those are period-end accruals approved the next morning, which may be an accepted practice worth documenting. Three entries above $10,000 have a single approval, and three have no support in Drive or NetSuite. Each exception in the workbook links to its approval log record and the file it was tested against, so a reviewer can follow the same path.

Sent for review. Nothing is filed until a reviewer signs off on the workpaper.
Reply…

How it works

  1. 1

    We turn your test procedures into institutional knowledge

    We sit with your team and write down how each control is actually tested: the population, the attributes, the thresholds, what counts as an exception. That becomes a test procedure the agent runs the same way every quarter.

  2. 2

    Upload the RCM and evidence

    Upload the RCM and the exports you already pull, or connect the source systems. Ask for the test the way you would brief a staff auditor. The agent gathers the evidence and performs the test.

  3. 3

    You review and sign off

    You get a workpaper with the population, the sample, the exceptions and an evidence index, every row tied to the record it was tested against. Ambiguous cases are flagged for your judgement, never auto-passed.

Judgement stays with your team

The agent drafts the workpaper. Your team makes the call on every exception, and nothing is filed until a reviewer signs off.

XLSJE-03_Q3_workpaper.xlsxDraft · awaiting reviewer
JE-2026-07-0912 · approved by its own preparerException
JE-2026-08-0331 · posted 9 minutes before approvalYour call
JE-2026-09-0044 · $27,900 with a single approvalException
JE-2026-08-1102 · support located after re-searchCleared
Nothing is filed until this is signed.Send backSign off

Every action is traceable

Every command, query and raw response is recorded and replayable. Each conclusion carries the file, field and record it was tested against.

Exception JE-2026-07-09124 records behind this conclusion
  1. Journal entryJE-2026-07-0912 · $18,400.00 · NetSuite
    Pulled 14:02:11View
  2. Approval recordPreparer R. Meyer · Approver R. Meyer · 04 Aug 17:02
    Pulled 14:02:16View
  3. Support searchFinance / Close / FY26 Q3 / JE support · not found
    Searched 14:02:20View
  4. ConclusionFail: preparer ≠ approver · dual approval · support attached
    Written 14:02:23View

Tested every week, not once a quarter

Once a test procedure is written, the agent runs it on every new record as it lands. Exceptions surface the week they happen instead of at quarter-end, and the workpaper is already drafted when your reviewers sit down.

JE-03 · Manual journal entry approvalWeekly
Q3 FY2613 runs · 1,284 entries tested · 3 exceptions
Wk 36 · 07 SepTesting 118 entriesRunning
Wk 35 · 31 Aug131 entries · 1 exceptionSigned off
Wk 34 · 24 Aug96 entries · 0 exceptionsSigned off
Wk 33 · 17 Aug142 entries · 2 exceptionsSigned off

Evidence retrieval from the systems you use

Connect the places evidence lives, or upload the exports you already pull for testing. The agent locates the right records, keeps the original responses as evidence, and only writes back to a system when you say so.

Google DriveGoogle SheetsOneDriveGmailOutlookSlackJiraWorkdayNetSuiteGitHub

Security

Audit evidence is sensitive by definition. Traack is built so your team can answer your own auditors' questions about where it went and who touched it.

traack

See it on one of your controls

Bring an anonymised RCM and an export. In thirty minutes you will have a results workbook and a clear view of what the agent can take off your plate this quarter.

Book a demo