Guide

Audit software for Chartered Accountant firms: 10 things to check before you buy

Questions that separate audit automation that holds up in review from a spreadsheet with a new name.

DeviSofts Engineering Team Updated 2 November 2026 8 min read

Audit software is easy to demonstrate and hard to judge. Every product can open a spreadsheet and draw a chart. What matters to a Chartered Accountant firm is whether the work it produces will stand up to a partner's review, a client's challenge and a regulator's inspection. These ten checks focus on that.

1. Does it test the whole population?

Sampling exists because people cannot check everything by hand. Software can. Ask whether every test runs on every transaction in scope, and whether the result lists each exception individually, with the document and the rule it failed. Testing 100% of invoices changes the auditor's job from picking a sample to judging the exceptions.

2. How does evidence come in, and can it be changed?

Client evidence arrives as Excel exports, PDFs, scanned paper and photos. Good software accepts whole folders, recognises the document type, removes duplicates and stores each file in a way that cannot be quietly altered. A cryptographic hash (such as SHA-256) recorded at intake lets you prove later that a file is exactly what the client gave you.

3. Can it read scanned documents offline?

Scanned invoices and delivery slips are where manual retyping eats days. Ask whether the software reads scans itself (OCR), whether that happens on your machines or on an outside service, and how it shows text it was not sure about. For confidential engagements, offline OCR avoids sending client documents anywhere.

4. Can your firm change the rules without new code?

Every engagement is a little different, and terms of reference change from year to year. Test parameters, tolerances, on/off switches, reason codes, document keywords, rates and even the wording of the certificate should be settings your team can change, with the change recorded. If every adjustment needs the vendor, you will wait for them at the worst possible moment.

5. Is every item classified with a reason?

A long list of "exceptions" is not a conclusion. Each transaction should end with a clear outcome (for example eligible, ineligible, unverified or unsubstantiated) and a reason code explaining why, so totals by outcome reconcile back to the population and reviewers can drill into any figure.

6. Does it enforce preparer and reviewer sign-off?

Review is the heart of audit quality. Each step of the audit programme should be signed off by a preparer and then by a different reviewer, and the software should refuse to let the same person do both. Roles (partner, manager, auditor, viewer) should control who can change what.

7. Is there a real audit trail?

Look for a field-level trail: who changed which value, from what, to what and when. When an auditor overrides a test result, the software should demand a written reason and keep that decision when the tests are run again. Without this, you cannot show an inspector how a conclusion was reached.

Test it: override one result in the demo, re-run all tests, then ask to see who changed it and why.

8. What does it hand you at the end?

The output is what the client and regulator see. Check for a certificate or report in an editable Word format, annexures in Excel (registers, exceptions, tests performed, an evidence index), and a consolidated report across entities or periods. Drafts with open items should be clearly marked as drafts.

9. Where does the data live, and how is it protected?

Ask where the database sits, whether the software needs the internet, and how passwords and database credentials are stored. An offline Windows application with the firm's own SQL Server keeps client files on the firm's machines. Check for account lockout, forced password changes for new users and one-click backups.

10. Can the vendor prove it works?

Ask how the software is tested. Automated tests, and a test run on data with known, deliberately planted exceptions, are strong evidence that the results are right. Ask how the methodology was built and who checked it, and agree in writing on setup, training and support before you start.

Questions to ask in the demo

  • Load a real (anonymised) client folder. How long until the first test results?
  • Show me one exception from the summary down to the source document.
  • Change a tolerance. Who can do that, and where is it recorded?
  • Try to sign off the same step as preparer and reviewer.
  • Disconnect the internet. Does everything still work, including OCR?
  • Generate the certificate and annexures for one period.

How SMAT Audit360 fits

Our SMAT Audit360 was built for price differential claim audits: it runs 22 automated tests on every claimed invoice, hashes and classifies evidence into 24 document types, reads scans with offline Windows OCR, keeps rules and certificate wording in editable settings, enforces preparer and reviewer sign-off, and runs fully offline on the firm's own SQL Server. The case study walks through a full engagement.

Questions

Quick answers

Can audit software test every invoice instead of a sample?
Yes, when the client's data can be read in a structured form. Software can run the same tests on the whole population and list every exception, which leaves auditors to judge the exceptions rather than pick a sample.
Is cloud audit software safe for client data?
It depends on the engagement. Some terms of reference and confidentiality agreements require client documents to stay on the firm's own machines; in that case an offline application with its own database is the safer choice. Ask where documents, OCR and backups are processed.
What should an audit trail record?
Who changed what and when, at field level; who prepared and who reviewed each step; and the written reason for every override of a test result. Decisions should survive when tests are run again.
DeviBot
Usually replies in a few seconds
Hey there 👋 I'm DeviBot. What's your name?