Fixed checks and automated review before sign-off

Which checks Achekam runs on draft financial statements, which findings are advice only, and what still rests on the reviewer and the partner.

By Achekam · 11 October 2026 · 5 min read · For Accountants & audit firms in South Africa, Nigeria, Kenya

"Which of these would actually stop me signing?" It is a fair question from a manager looking at a draft set of annual financial statements with a list of findings beside it. Some checks in any system are arithmetic and cannot be argued with. Others are suggestions. If the team cannot tell the two apart, either the suggestions get ignored or the arithmetic gets treated as optional.

This article separates them for Achekam: the fixed checks, the automated review, the AI material that sits in a draft, and the single control that does stand between a draft and a signature.

Two layers that are easy to confuse

Achekam checks a draft in two different ways, and they do different jobs.

The Validation & Math Engine runs fixed checks that involve no AI, such as whether double entry holds and whether the notes tie to the statements. These are rules written in code. Make sure they pass every time a draft is rebuilt: in Achekam every figure in the statements is calculated in code from the mapped trial balance, so a broken tie is more likely to point to a mapping problem or a figure someone typed in by hand than to an arithmetic slip.

The automated review is something staff choose to run on a draft. It produces a list of findings, and a person accepts, resolves or dismisses each one. It is advice only: an open finding does not block signing.

Keep the names distinct in your firm's language. "The validation failed" and "the review raised something" should mean different things to everyone on the job.

Working the findings list

Accept, resolve and dismiss are three different decisions, and your review procedure should say what each one means in your firm. A workable convention:

• Accept when the finding is right and the draft has been, or will be, changed because of it.
• Resolve when the issue has been dealt with, for example a disclosure was added or a mapping corrected, and you can point to where.
• Dismiss when the finding does not apply to this entity, and the reason belongs on the file.

The danger with any advisory list is that it is cleared in a hurry the day before the client meeting. Because findings do not block signing, the only thing that makes them matter is the reviewer's habit of reading what was dismissed and why. Where a dismissed finding concerns a disclosure, the reason is often a materiality judgement, and the IFRS Foundation's practice statement on making materiality judgements gives non-mandatory guidance, a four-step process for making those judgements when preparing IFRS financial statements.

What sits in the draft that a person has not yet looked at

Two kinds of AI material can be in a draft before anyone has reviewed it, and the reviewer needs to know where they are.

Mapping proposals. When the Studio builds a draft, AI mapping suggestions with a confidence of 0.5 or more go into the draft at once, and are listed as proposals that a person can accept, move or set aside. The research behind this article found no rule that stops signing while such proposals are still open. The workbench (Canvas, then "Mapping") shows each account's line and note and lets staff move accounts by hand.

Narrative text. Where an AI key is set up, AI writes narrative such as the nature of the business and note wording. It is checked in code and labelled "AI draft - review before issue". There is no separate accept step for each paragraph. Without a key, standard wording is used instead.

A third item to search for is not AI at all: facts that the trial balance does not hold are left in the draft as "[to be completed]". A reviewer who searches for that marker before reading for sense saves an embarrassing conversation with the client.

The control that does hold

What Achekam enforces at the end is about people, not findings. Each set of statements names a preparer, a reviewer and an engagement partner. Only a user with a partner role can sign, and the person recorded as preparer or reviewer cannot issue the set.

So the safeguards against an unreviewed AI proposal or an unread finding are the review stages and the partner's signature. That places the weight where professional standards already place it: on the people who review and sign.

A pre-signing routine for the reviewer

Before the draft goes to the partner, the reviewer can work through the same short routine on every job:

1. Rebuild the draft and confirm the fixed validation checks pass.
2. Open the mapping proposals list and clear every proposal: accept it, move the account, or set it aside.
3. Run the automated review and read every dismissed finding, not only the open ones.
4. Search the draft for "[to be completed]" and for the "AI draft - review before issue" label, and read each labelled passage against what you know of the client.
5. Confirm the preparer, reviewer and engagement partner named on the set are the right people.

On audits and reviews where the firm has named an engagement quality reviewer, that person's work comes on top of this routine; see naming an engagement quality reviewer. Matters that needed a consultation should also be closed before signing, because, as explained in quality registers and consultations, an open consultation does not stop an audit report from being approved or signed.

Where to start

Take one recent draft, run the automated review on it and sit with the preparer while you work through the findings and the open mapping proposals together. Agree what "dismiss" means in your firm, write it into your review checklist, and use the same checklist on the next set you sign.

General guidance only, not advice on your own circumstances. Tax and payroll rules change: check the current position with the authority or your adviser.