Working observations from current engagements
Practical, long-form, and unfashionably specific. Written between engagements rather than during them.
-
What AuditEngine actually generates, and what it deliberately doesn’t
Workpaper generation is a structured-output problem. Judgment is not. The split is what makes the architecture deployable in a regulated environment, and what stops it from drifting into territory that is not a tooling problem.
-
The FRC’s recent inspection findings, read in slow motion
Files do not fail because the auditor reached the wrong conclusion. They fail because the file does not visibly contain the reasoning that produced the right one. That is a different problem with a different fix.
-
IFRS 9 ECL governance: where bank files quietly fail
Bank ECL files fail in four recurring ways: staging triggers written but not enforced, scenario weights governed but not documented, post-model adjustments that grow without challenge, and disclosure narratives inconsistent with the underlying inputs.
-
Going-concern files that survive forty minutes of FCA reading
A good going-concern file is six things, in order: the period assessed, the cash and liquidity baseline, the regulatory and capital baseline, the management forecast, the stress, and the conclusion. Each section is two to four pages. The file is forty to fifty pages total.
-
Anatomy of a first statutory audit for a UK fintech
A first audit turns on six topics in order of weight: revenue recognition under IFRS 15, safeguarding compliance evidence, classification of customer monies, going concern with equity rounds in scope, share-based payments, and intercompany debt.
-
JSOX component assessments without the four-week binder
Japanese parent groups still need component certification. The deliverable is mandatory. The format is not. A well-structured RACM and a tight aggregation memorandum replace fifteen percent of the binder volume without losing a single piece of evidence.
-
From ten days to five: a close-compression case study
The two reliable levers are sub-certification cascades and reconciliation libraries. Both are unglamorous. Both shift the bottleneck from accounting to review, which is exactly the right place for it.
-
Building societies and the quiet narrative trap in FRS 102
Mutuals produce three narrative artefacts each year: Annual Review, Pillar 3, and strategic report. They should agree and frequently do not. The fix is a single workpaper, drafted by audit, that ties every assertion across the three documents to a single source.
No notes in this category yet. More to follow.