The groundwork, done. The judgement, yours.
Scrutive AI is built into the audit workflow rather than bolted on as a chat window. It reads the same evidence your auditors read and produces drafts in the formats they already sign.
A comprehensive program, in minutes, for any scope.
Scrutive first maps every area the scope must cover, from governance and segregation of duties to systems, compliance and fraud, then drafts each area in depth.
- 5 to 8 risks per area, each with a rating and the control expected
- 6 to 10 test steps covering design, operating effectiveness and data analytics
- Evidence to obtain, samples sized to control frequency, red flags
- Tick what to keep; add to the audit and, optionally, the control register
In our HR audit trial run: 11 areas and 56 risks and controls in about 90 seconds.
Every control point checked against the evidence.
When an owner says a control is in place, Scrutive reads their files and links against each point of the control and tells the verifier what is covered, what is partly covered and what is missing.
- PDFs, scans and photos, Word, Excel and PowerPoint, SharePoint links
- Concerns flagged, such as dates, signatures or self-approval
- A ready return note listing exactly what to send
Samples tested item by item, exceptions surfaced.
Attach the sample to a work paper and Scrutive tests each item against the control and the planned test, records the attributes checked, lists exceptions and drafts the test performed and observation.
- Pass, exception or not applicable for every item
- Conclusion and follow-up questions for the auditor
- Nothing written into the work paper until the auditor chooses
Designed for functions that audit AI, too.
Internal audit cannot rely on a black box. Scrutive's AI is bounded, recorded and always subordinate to professional judgement.
Human in command
AI output is a draft. Work papers holding an AI draft cannot be sent for review until the auditor replaces it with the test performed.
Recorded
Each result shows what was read, what could not be read, who ran the check and when — all in the audit trail.
Controlled
Switched on by a master, with a daily request limit. If a limit or the provider's quota is reached, AI pauses and resumes the next day.
Your data stance
Choose AI settings leadership is comfortable with — including options where your files are not used to train models. The same notice appears wherever AI is turned on.
Honest about limits
The AI says when evidence is not enough to conclude, and lists files it could not read, rather than guessing.
Resilient
If one AI service is busy, another completes the same request so your fieldwork is not held up.
Watch Scrutive build a program for one of your audits.
Bring a scope. In the briefing we draft the full program live and test a sample of your evidence.