How the leader track works
Six sessions, thinking mode, zero code. This track is for the people who decide - not the people who build. Tonight sets the map; a2 scores your use-cases, a3 covers trust and governance, a4 builds the team, a5 proves the ROI, and a6 sequences the 24-month roadmap. Every session ends the same way: a short list of questions to ask your team and your vendors, so you walk out able to lead the conversation, not just sit in it.
The state of play: past hype, past fear 5 min live
Two failure modes stall a finance function: the leader who believes the demo and the leader who bans the tool. Both are expensive. The adoption is already happening - 59% of finance functions now use some form of AI, and nearly 4 in 5 CFOs say they will deploy generative AI within 24 months. Your job is not to be an enthusiast or a sceptic. It is to know precisely what lands where.
LiveThe hype cycle in finance - and how to stand in the middle of it3 min▶
Every finance leader is being pitched two stories at once: "AI will replace your team" and "AI is a toy that makes things up". Both are wrong, and both are being sold to you by someone with an incentive. The truth is narrower and more useful: today's AI is a superb drafting and analysis assistant over data you supply, and a poor autonomous decision-maker. Judge every pitch against that line.
- What is real today. Drafting commentary and memos, analysis over numbers you hand it, document Q&A, extraction from invoices and contracts, first-pass anomaly flags. These save real hours this quarter.
- What is still vaporware. The "fully autonomous CFO", closing the books untouched, or an agent that files to a regulator on its own. Impressive on stage, unaccountable in an audit.
- The tell. If a demo hides the human review step, it is selling the fantasy. Real finance AI is loud about where the person signs off.
LiveWhat is real in 2026 - the can / can't line3 min▶
Builder session b1 spends an hour proving this to practitioners with live experiments; here is the leader's version. An AI is a language engine, not a calculator or a database. That single fact draws a clean line your whole function can use.
| Trust it to (with your data + a check) | Do not trust it to (unaided) |
|---|---|
| Draft variance commentary and board narrative | Recall a specific figure from memory |
| Analyse and explain numbers you supply | Do multi-step arithmetic silently |
| Answer questions over a document you attach | Cite a source it was never given |
| Extract fields from invoices and filings | Be your system of record |
The left column is where your quick wins live and where the builder track spends its time. The right column is why governance (a3) exists. 2026 curricula such as Chicago Booth now teach agentic and multimodal AI as first-class topics - real, but firmly in the "govern before you trust" camp, which is exactly what a6 covers.
The three eras of finance AI 4 min live
Every AI pitch you will ever be shown fits into one of three eras. Naming them turns a confusing market into a simple question: which era is this, and are we ready to govern it? Most finance teams today are living in eras one and two. Era three is emerging fast - and must never be adopted before the governance for it exists.
LiveAssistive, augmented, agentic - and where your team actually is2 min▶
- Era 1 - Assistive. The AI drafts and summarizes; a person reads and uses every output. Think "faster first draft of the commentary". Low risk, immediate value, start here.
- Era 2 - Augmented. Analysis copilots reason over the data you supply - explaining a variance, stress-testing a forecast, surfacing anomalies. Still a human decision at the end, but the AI does more of the thinking.
- Era 3 - Agentic. The AI runs multi-step workflows on its own - pulling data, reconciling, drafting, routing - with approval gates a human controls. Real and coming, but it multiplies both the value and the blast radius. Governance is not optional here.
The honest answer for most finance functions in 2026: you are somewhere in eras one and two, and the smart move is to get excellent there before climbing. The teams that get burned are the ones sold an era-three agent before they had era-one habits.
The leader's real job in this shift 4 min live
You do not need to write a prompt to lead this well - although it helps to understand what your team is doing. Your job is the set of decisions no tool can make for you. This track is built around exactly five of them, one per remaining session.
LiveThe five decisions only you can make2 min▶
- Pick the bets (a2). Score every candidate use-case on value and risk, and choose the first few deliberately. Most AI programs fail on picking, not building.
- Govern it (a3). Model risk, explainability, audit trail. Draw the line between "assist" and "decide" and make it stick.
- Build the team (a4). Roles, skills, and the change management that turns fear and free-for-all into a shared way of working.
- Prove the ROI (a5). Measure value honestly, move a pilot to scale, and decide buy versus build with numbers a board will accept.
- Sequence the roadmap (a6). A 24-month plan that earns trust before it climbs eras - not a wishlist of demos.
LiveThe two rails that run through everything2 min▶
Before any use-case, any tool, any era, two rails hold. Say them out loud to your team, because they are what separate a finance function that uses AI well from a headline about one that used it badly.
- Rail 1 - the privacy line. Never paste material non-public information, client PII, or unpublished financials into a consumer AI tool. This is securities law, data-protection law (PDPA, GDPR), and your firm's confidentiality obligations at once. Real data goes only into enterprise or local tools.
- Rail 2 - AI assists, the licensed human decides and owns it. The model drafts, suggests, and flags; a qualified person reviews and owns every number, filing, and decision that leaves the building. Nothing in this course is investment, tax, accounting, or legal advice, and no workflow makes an automated decision about a person's money.
Samsung's ban. Engineers pasted confidential source code into a consumer chatbot to debug it; the company banned generative AI on internal machines within weeks. The finance equivalent - pasting a pre-release earnings table to "help write the commentary" - is worse, because it can also be insider-information mishandling. Rail one exists so no one on your team ever becomes that story.
Map your top 3 time-sinks ◐ 15 min · with your team
No code, just a conversation and a whiteboard. This is the raw material a2 turns into a scored shortlist of bets.
With one or two people who know the work, list your function's three biggest recurring time-sinks - the tasks that eat hours every month (the close narrative, the board pack, invoice matching, cash reporting, whatever it is for you).
For each, note which era it fits - is it drafting-and-analysis (era 1-2, safe to pilot) or decision-making (govern first)? Use the value map from Part 1 as your reference.
For each, flag the data sensitivity: would rail one let you use a consumer tool on it, or does it need an enterprise or local path?
Bring the three to session a2, where you will score them on value and risk and pick your first deliberate bet. Do not try to solve them tonight - just name them clearly.
Official sources covered
This track teaches from public finance-AI curricula and verified industry data. Certificates, graded assessments, and paid tooling stay with the official programs - all noted honestly. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · A vendor demos an AI that "closes the books and files with no human touch". What is the realistic read in 2026?
What is real is assistive and augmented work over your data. The "autonomous CFO" hides the review step - and accountability can't be outsourced to a model.
2 · Where are most finance teams honestly sitting on the three-eras model?
Adoption is real (59% already use some AI) but concentrated in drafting and analysis. Era-three agents multiply the blast radius, so earn era-one habits first.
3 · In this shift, what is the finance leader's actual job?
You don't need to code. The value is in the five decisions that carry real accountability - which is exactly what a2 through a6 walk through.