Approach
How the check works
- 01
You send the draft.
Before we receive anything, we agree on confidentiality and insider procedures. You send the near-final draft, in Finnish and English, through an agreed secure channel. We need no access to your systems.
- 02
We check it.
A set of narrow AI agents works through the draft in parallel, each on its own check. Figures are extracted from the document by software. We collect your earlier releases ourselves from public sources. Before delivery, a person verifies every finding we raise against the original page.
- 03
You receive the findings and decide.
Findings arrive within 48 hours, with page references, ordered by what should be resolved before publication. Each needs only one of three answers: fix it, keep it deliberately, or not valid. We do not publish, send or change anything.
What we check
Figures
Every figure is recalculated: totals, quarters derived from half-year figures, comparison columns, and every place the same figure appears in the report.
Language versions
The Finnish and English versions are compared claim by claim: does the translation say the same thing with the same degree of certainty, and does each caveat appear in both.
Disclosure history
The draft is compared with what you have published before: comparative figures, earlier estimates and targets, and risks you have named previously.
Key figure definitions
Definitions and reconciliations of alternative performance measures, period by period: has a definition changed, was the change disclosed, and can a reader derive the headline figure from the report's own numbers. In 2025, European national enforcers examined the presentation of alternative performance measures 434 times after publication, and 16 per cent of these examinations led to enforcement action (ESMA).
Framing
Does the narrative tell the same story as the figures: the CEO's review against your own tables, one-off items treated with the same discipline in both directions, and what the highlights leave unsaid.
Requirements and hygiene
We also check the items IAS 34 requires and the technical hygiene of the document, such as draft stamps and unfilled template fields.
What you receive
A list of findings. It is working material for you, not an opinion and not an assessment of anyone's work.
For each finding we state what we observed, where it is (page and passage), why it matters, and which earlier release it relates to. What should be resolved before publication comes first.
We also record what has been done well, using the same standard.
Illustrative example, not a client finding
01 Margin presentation
Check required- Observation
- Two pages use different margin figures for the same period.
- Source
- Draft version, pages 4 and 9.
- Why it matters
- Readers may quote either figure. The difference may indicate an outdated table.
- Next step
- Confirm the intended figure and update the other reference.
What our checks have found so far
We have developed the method on the published material of five companies: three Finnish listed companies, an unlisted bond issuer and a European banking group. We do not name them.
check runs on the published material of five companies
findings, of which the agents classified 53 as to be resolved before publication
key findings verified against the original page. None was overturned, and six were refined
Three types of finding recurred in four of the five companies
A key figure definition changed without explanation, an earlier estimate or target was not followed up or changed quietly, and an item required by regulation was missing.
Examples of findings
- The same key figure published with several values and no reconciliation.
- A headline growth figure that cannot be derived from the report's own numbers.
- A caveat in the headline missing from the English version.
- Comparative figures recalculated on a new basis without saying so.
- A one-off cost adjusted out, while the matching one-off gain was not.
- A statement in the CEO's review that contradicts the report's own figures.
What the numbers do not tell you
Findings are items to check, not established errors. The findings verified against the original page were the key ones, not a random sample, so the result does not show the share of false alarms among all findings. The checks were run on published material.
What we are not
We are not your auditor.
The check does not replace the audit and does not shift responsibility. Our findings are material you can take up with your auditor.
We publish nothing.
We have no access to your publishing systems, and we do not need it.
We do not assess people.
Findings concern the text, not the authors.
We are not a software project.
You need no new tools: you send the draft and receive the findings.
