The AI Learning Hub Journal

What the Inspector Will Ask

The file is written under deadline and read at leisure, scepticallythat asymmetry has always disciplined audit work — AI hands the sceptical reader a new object, not a new disciplineassembled under deadlineby the team that wrote itarchiveclosed — nothing can beadded to it afterwardsyears passread by reviewers and inspectorswhose duty is scepticism of its authorswhatever was relied on — a junior’s schedule, a specialist’s valuation, an analytics routine or a tool — unexplained reliance is a findingTHE PREDICTABLE FINDINGS WRITE THEMSELVESreliance the file cannot explainwhy this tool, on this evidence, wasreasonable — the file never saysevidence that is a screenshotoutput existed; what went in and whatanyone did about it are not recordeda conclusion nobody tied backoutput fed the opinion with no sourcedocument opened behind itnone is a new category of finding — the old questions, attached to the newest thing in the fileTHE WORKPAPER QUESTIONS — ANSWERABLE IN ADVANCE OR NOT AT ALLwhat did the tool do —which procedures didits output feed?what went in — was itwithin what the firmpermits to enter?what came out — andwhere is that outputheld now?what did the team do —flags followed to source,and the rest resolved?why was reliancereasonable — what didthe firm know of it?the hardest follow-up concerns the misstatement the tool did not flag — the only defence is procedures that did not stop at its flagsTHE FIRM ABOVE THE FILE — QUALITY MANAGEMENT (ISQM TERRITORY) NOW GOVERNS THE TOOLSONE FILE“why did this team rely on this tool?”× every teamTHE FIRM’S SYSTEM“how does the firm know, for every team, that the answer is a good one?”a tool adopted faster than the firm noticed is an unidentified risk to engagement quality — a gap here predicts findings in files not yet inspectedTHE ANSWERS EXIST IN ADVANCE OR NOT AT ALLa file cannot be explained backwards three years on — and neither can a firm’s decision to permit the tool
The inspector’s AI questions are the profession’s old questions attached to a new object — answerable in advance or not at all.

The File Is Read Years Later, Sceptically

The audit file is an unusual work product: assembled under deadline, then read at leisure, years later, by reviewers and inspectors whose professional duty is scepticism about the team that wrote it. That asymmetry has always disciplined audit work, and AI does not change it — it hands the sceptical reader a new object. The predictable findings write themselves: reliance on a tool the file cannot explain; evidence that turns out to be a screenshot of a results screen; a conclusion resting on output nobody tied back to a source document. None of these is a new category of finding. Inspectors have always asked why the team relied on what it relied on, whether the thing relied on was a junior's schedule, a specialist's valuation or an analytics routine. The questions are old; the tool is merely the newest thing they attach to.

  • The file is written under deadline and read years later by people paid to be sceptical of its authors
  • Reliance the file cannot explain is a finding, whatever was relied on — junior, specialist or tool
  • A screenshot records that output existed, not what produced it or what anyone did about it
  • The inspector's AI questions are the profession's old questions attached to a new object

The Questions, Concretely

The questions themselves are worth rehearsing, because they are answerable in advance or not at all. What did the tool do on this engagement — which procedures did its output feed? What went into it, and was that within what the firm permits? What came out, and where is it now? What did the team do about the output — which flags were followed to source documents, and what happened to the rest? And why was reliance reasonable — what did the team, or the firm on its behalf, actually know about how the tool behaves? These are the workpaper questions from earlier in this course, unchanged by the setting. The uncomfortable follow-up is the one about the misstatement the tool did not flag: the only defensible answer is a file showing that the team's procedures did not stop where the tool's flags stopped.

  • What the tool did, what went in, what came out, what the team did about it, and why reliance was reasonable
  • The answers exist in advance or not at all — a file cannot be explained backwards three years on
  • Flags followed to source documents are evidence; flags noted and filed are a list of open questions
  • The hardest question concerns what the tool missed, and the answer is procedures that did not stop at its flags

The Firm Above the File

Above the engagement sits the firm, and above the file sits the firm's system of quality management. The quality-management standards — ISQM 1 is the international instrument — ask, in substance, that a firm identify risks to engagement quality and design responses to them. A tool that engagement teams adopted faster than the firm noticed is precisely such a risk: nobody assessed it, nobody monitors it, nobody trained anyone on its failure modes, and yet audit evidence now passes through it. The firm's system is inspection territory as well as the individual file, because a gap at firm level predicts findings in files not yet inspected. The firm-level questions are the workpaper question at scale: not why did this team rely on this tool, but how does the firm know, for every team, that the answer is a good one.

  • Firm-level quality management is meant to govern how audits are performed, and tools are now part of how
  • A tool adopted faster than the firm noticed is an unidentified, unanswered risk to engagement quality
  • A gap in the firm's system predicts findings in files that have not been inspected yet
  • The firm-scale question: how does the firm know every team's reliance would survive the same scrutiny

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.