Building draft statements from a mapped trial balance
What Achekam's Studio builds when a firm drafts annual financial statements from a mapped trial balance, which checks run, and what is left to people.
By Achekam · · 4 min read · For Accountants & audit firms in South Africa, Nigeria, Kenya
Most of the hours in a compilation or audit file are not spent on judgement. They go on carrying balances from a trial balance into a statement of financial position, adding up notes, making the cash flow agree and retyping the same policy paragraphs as last year. Software can take that work away. The question a partner should ask is what it takes away, and what it leaves for people to check.
This article walks through what Achekam's Financials Studio builds from a mapped trial balance, where the figures come from, what checks run on the draft, and the parts of the set that a reviewer still needs to read line by line.
From approved mappings to a draft set
The Studio's starting point is a trial balance whose lines have been mapped to statement lines. Achekam suggests where each line belongs, from the file's own mapping column, earlier mappings, AI or keyword rules, and no suggestion becomes an approved mapping until a person approves it.
There is one nuance worth knowing before the first build. When the Studio builds a draft, AI mapping suggestions with a confidence of 0.5 or more are placed in the draft straight away and listed as proposals that a person can accept, move or set aside. So a first draft may already show accounts in lines that nobody has approved yet. That is useful for speed, but the proposals list has to be worked through before the draft is treated as anyone's view.
From those mappings the Studio builds a full draft set:
- the directors' report;
- the four primary statements;
- accounting policies and notes;
- a detailed income statement.
What a complete set must contain is set by the reporting framework. The IFRS Foundation's summary of IAS 1 describes the overall requirements for the presentation, structure and minimum content of financial statements.
Every figure calculated in code
Every figure in the draft is calculated in code from the mapped balances. Nothing numerical is written by an AI model. Where an AI key is set up, AI may write narrative text such as the nature of the business or note wording, and that text is labelled "AI draft - review before issue". Without a key, standard wording is used.
The practical consequence is that a wrong number in the draft almost always traces back to the trial balance or a mapping, not to the drafting. When a figure looks wrong, go to the mapping first.
The practitioner's report follows the engagement level
The practitioner's report in the set follows the level chosen for the engagement: audit, independent review, compilation or none. Each level answers to its own standard: ISA 700 for an audit opinion, ISRE 2400 (Revised) for an independent review and ISRS 4410 (Revised) for a compilation, so read the report in the draft against the standard the engagement is actually performed under. Choose the level before the first build, because it changes what the reader is told about the work done.
Edited blocks survive a rebuild
Drafts are rebuilt many times: a late journal, a corrected mapping, an adjustment agreed with the client. A block of text that a person has edited is kept when the draft is rebuilt, so a carefully reworded going-concern paragraph does not disappear because a balance changed.
The reverse also matters. Because edited blocks are kept, an edited paragraph that mentions a figure will not update itself if the figure changes. Search edited blocks for numbers or descriptions that the latest rebuild may have made stale.
Two kinds of check
Two separate sets of checks look at the draft.
| Check | What it is | Effect |
|---|---|---|
| Validation & Math Engine | Fixed, non-AI checks such as double entry and note ties | Shows what does not agree |
| Automated review | Findings listed for a person to accept, resolve or dismiss | Advice only; does not block signing |
The automated review is a second pair of eyes, not a gate. Its findings are advice, and nothing in it stops a partner signing. Treat each finding as a review note: decide it, and leave a reason when you dismiss one.
What a reviewer should still read line by line
Achekam carries the arithmetic; people carry the judgement. Before the set goes to the partner, a reviewer should read:
- The proposals list. No rule was found that stops signing while AI mapping proposals are still open. The safeguards are the review stages and the partner's signature, so clear every proposal deliberately.
- Every [to be completed] marker. Facts the trial balance does not hold are left marked rather than invented. Each one is a question for the client.
- Policy wording. The standard policies are based on IFRS for SMEs and labelled for review before issue. Check they match the framework and the client.
- AI narrative. There is no separate accept step per paragraph, so read each labelled paragraph against what the client really does.
- Edited blocks. Confirm they still agree with the rebuilt figures.
- Automated review findings. Accept, resolve or dismiss each one, with reasons.
Each set 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 review stage is a real step and not a formality.
Getting the first draft right
Approve the mappings, pick the engagement level and only then build. The first draft will be closer to final, and the reviewer's time goes on judgement rather than on chasing misplaced accounts. The wording that comes with the draft is explained in accounting policies and notes: what is filled in, and the AI's narrower role in what the AI writes in statements, and what it does not.
General guidance only, not advice on your own circumstances. Tax and payroll rules change: check the current position with the authority or your adviser.
