AI · IFRS · PROFESSIONAL JUDGMENT

AI knows IFRS, so why can’t every answer be trusted?

Knowing a standard is not the same as building a professional conclusion. A practical example shows where generic answers fail and what methodology adds.

Knowing the standard is necessary, but it is only the beginning. A professional answer is not measured by how many standards or paragraph numbers it mentions, but by whether it identifies the facts that change the conclusion, tests the counterargument, and states the limits of what can be concluded.

Example: a SaaS contract commission

A company pays an 8% commission to obtain a new annual subscription contract. Average customer life is four years and renewal commission is 2%. The apparent question is simple: capitalize or expense, and over what period should any asset be amortized?

A generic answer may say: “Capitalize it because it is an incremental cost of obtaining the contract under IFRS 15, then amortize it over the contract term.” The headline may sound right while the conclusion remains incomplete or wrong.

The facts that decide the conclusion

  • Would the commission genuinely not be paid unless the contract is obtained?
  • Does the contract renew automatically, and what evidence supports expected renewals?
  • Is the renewal commission commensurate with the initial commission? If not, the initial commission may benefit renewal contracts.
  • Is the four-year life based on an appropriate cohort of comparable customers and contracts?
  • Do different products, pricing structures, or sales channels change the benefit pattern?

Without these facts, “four years” is not an answer, and the one-year contract term is not necessarily the amortization period. This is the difference between recalling a rule and applying it.

Where does AI commonly fail?

1. It answers the written question, not the real accounting issue

It may focus on capitalization and miss the harder issues: amortization period, impairment, or renewal commission economics.

2. It silently fills missing facts

Confident wording may conceal assumed renewals or assumed commensurate commissions.

3. It mixes requirements, practice, and conclusion

What the standard requires, what peers do, and what management concludes are three different layers. A professional answer separates them.

4. A citation may look precise without being accurate

A paragraph number is not reliable merely because it is stated confidently. Material issues require verification against the actual text and effective amendments.

What does methodology add?

A sound methodology forces the analyst to identify the governing issue, request missing facts, separate facts from assumptions, build the strongest case for and against the treatment, and issue a clearly conditional conclusion.

This is the purpose of IFRS Expert Skill: not to add IFRS knowledge to Claude, but to impose a more disciplined workflow that knows when to answer, when to request facts, and when official text must be reviewed directly.