Guides for preparers and Reviewers who use AI tools on client folders: cited figures, checks and workpapers.
How a Reviewer sorts AI-drafted rows into tiers, reads every top-tier row and a stated sample of the rest, and widens the check when a sampled row fails.
A two-week practice plan to train a new Preparer to check AI drafts: synthetic Doe drill files with seeded errors, a drill log, and a senior's sign-off on each drill.
How a manager scores client files on four workload facts to pick which go through an AI-drafting pass first and which stay all by hand.
Client folder setup before an AI pass: rename every file, set aside duplicates, mark superseded versions, and build a one-page file index a person checks.
How a Reviewer writes review notes on AI-drafted work (what, where, expected fix, cleared-by) and how the Preparer clears and re-submits each one.
How a Preparer or Reviewer ties out AI-extracted rows to a document's own totals: control totals, page subtotals, and row count against item count.
Claude Code and Codex side by side for a CPA or EA firm's small internal scripts: instruction files, review, undo, ask-first settings and surfaces, in each vendor's words.
How a CPA or EA firm can use Codex, OpenAI's coding agent, to build small read-only checker scripts on Doe test files, with a person reading every diff.
How a CPA or EA firm sorts AI output into three kinds, writes the line task by task, handles a draft that crosses it, and records who moves it.
Build an AI instruction for tax-document work from six checkable parts, change it one part at a time with a change log, and version and retire it on purpose.
The full path from a client folder to an entry list the Reviewer can trace: what each step passes on, the gate before each hand-off, and what reopens.
How a CPA or EA firm sets ChatGPT's Chat and Work modes per task, what goes in the folder it can modify, and which plans OpenAI's training default covers.