Ten finance AI implementations on two pages, in the order that earns the next one.
Each entry carries the operating mode, the difficulty, a deploy estimate, the impact and the risk posture it runs under. Every one keeps a named human on the number; that is the methodology, not a caveat.
Three of the ten in full, and all ten named.
Eight run agent-led with a review step, two run AI-assisted and produce questions rather than answers. None of them decides anything. Across the ten, impact runs 3/5 to 5/5, and the three below are the only three I rate 5/5: the week-one anchor, the commentary build it feeds, and the one with the most to lose.
Variance InvestigatorAgent-led + review
Actual vs budget per line, drivers split known vs UNEXPLAINED, and the explanation re-adds to the total.
Medium · about 3 hours · impact 5/5 · analysis only, finance verifies
Month-End Commentary BuilderAgent-led + review
The verified variance pack in, draft commentary out, every claim tied to a pack line.
Low · about 2 hours · impact 5/5 · finance verifies before circulation
Reconciliation Exception AnalystAgent-led + review
Subledger-to-GL tie-out, exceptions isolated, aged and packaged. It never certifies a rec.
Medium · about half a day · impact 5/5 · sign-off stays human, always
The other seven, by name
CFO Morning Brief · Forecast Challenger · Management Reporting Assistant · Budget Assumption Reviewer · Expense Anomaly Analyst · Finance BP Meeting Prep · Board/CFO Pack Quality Challenger
On the card each of those carries the same five fields as the three above: operating mode, difficulty, deploy estimate, impact and risk posture. That is what the download is for, and it is the whole reason this page does not just reprint the card.
What the deploy estimates assume
Estimated build time for one experienced finance analyst or admin following an implementation guide, and excluding your organization's own prep: extract shapes, budget file hygiene, materiality thresholds. The prep is usually the longer half, and the card says so.
The first ten stop where a call moves money or attests to a number.
Journal approval. Payment release. Reconciliation sign-off. Forecast approval. Credit decisions. Anything where AI would make, rather than prepare or challenge, a decision of that kind.
These are not late-roadmap items. They are the never-automate list, and the gates around them are the product of a finance function, not overhead on it.
categories are named on the card as never automate, not deferred. If a finance-AI deck you are shown has fewer prohibitions than this free card, ask which control it deleted.
Ordered so the week-one wins buy the credibility for the week-four ones.
The sequence is the point. A finance team that opens with the hardest build spends its goodwill before it has shown a single number re-add.
Week 1, the anchorTwo builds
One build has to prove the whole idea before anything else earns attention: an explanation that re-adds to the total, or names what it cannot explain.
Week 2, the daily habitsTwo builds
The ones people touch every morning, so the layer stops being a project and becomes a habit.
Weeks 3 and 4, the trust buildersSix builds
The reporting pack, the tie-outs and the review passes, in the order that survives an audit conversation. The card names which build goes in which week.
Free. Pay $0 at checkout and the PDF lands in your library.
Licensed, not sold. Free resources ride the same License & Terms as everything else here (Article 3.1); the full notice is at the foot of this page.
Practitioner guidance, not accounting, audit, tax, legal, or compliance advice. AI prepares; your people decide.
Two pages, PDF. Pay what you want if you like it, starting at $0.
The card is the sequence. The Stack is the build.
The card is the two-page table and the never-automate list, nothing more. The Stack is where all ten arrive as tenant-tested agent builds with their fixtures and answer keys, so you can grade your own runs instead of trusting mine.