Hi Reader

If you are relatively new to my newsletter, maybe you don't know that much about me, so here is the 20 seconds intro. Also follow me on LinkedIn for daily tips on AI for Finance.

I'm Christian Martinez, Finance Senior Manager, AI for Finance Professor, and published author of Smart Finance: Leveraging AI for Enhanced Financial Planning and Analysis.

I've trained over 40,000 CFOs, Fractional CFOs, and senior finance leaders in AI for Finance through AI Finance Club, AI Finance Accelerator, LinkedIn Learning, and YouTube. I've spoken at the World Finance Forum, World Summit AI, Finance Digitalization Forum, and the EMEA FP&A Summit.

This week: Claude Cowork for FP&A and Finance teams + 6 Weeks Claude Adoption Plan

If you've been waiting for AI that can actually sit inside your close process, not just answer questions about it, this is the one to pay attention to.

Claude Cowork takes Claude's agentic capabilities (the same architecture behind Claude Code) and brings them to finance work. No terminal. No scripting. You describe the outcome, step away, and come back to a finished deliverable.

Here is my full and interactive guide.

What that looks like in practice

Instead of this:

You get this:

Claude plans it, breaks it into subtasks, runs it in an isolated environment, and delivers the workpaper — while you're in a meeting.

Why FP&A and Accounting teams should care

Close and reconciliation — Point Claude at your bank statement and GL export. It categorizes every reconciling item as timing, error, or unrecorded — and gives you the adjusted balance.

Journal entries — Draft accruals, depreciation entries, and deferred revenue adjustments — complete with supporting schedules and audit-trail descriptions.

Variance analysis and reporting — Decompose a revenue variance into price, volume, and mix drivers, and get a leadership-ready narrative — not just a spreadsheet.

Audit and controls — Pull a SOX sample, build a testing workpaper, and classify control exceptions by deficiency severity.

Data prep — Feed it messy ERP, POS, and CRM exports — mixed date formats, currency symbols, duplicate order IDs and all — and get back one clean, forecast-ready dataset.

Three prompts to try this week

1. "Reconcile this month's bank statement against the GL export. Categorize every reconciling item as timing, error, or unrecorded, and give me a summary table with the adjusted balance."
2. "Decompose the revenue variance vs. budget into price, volume, and mix drivers using this data, and write a two-paragraph narrative I can hand to leadership."
3. "Prepare the month-end accrual journal entry for unbilled AP based on this PO log, including the debit/credit lines and a supporting schedule."

One thing to know before you use it

Cowork has real risks given its file and internet access — this matters more in finance, where the files involved are sensitive. Use Manual approval mode for anything touching live GL, banking, or journal-posting systems. Save "Auto" or "Skip" for read-only reconciliation and analysis work you fully trust.

Do you want your Finance and FP&A team to adopt Claude?

Here is my full 6-week plan.

When to use it?

Imagine that your CFO just approved 200 Claude licenses.

You finally convince them that Claude has great potential for finance teams and that's great! But now how do you make sure your team actually knows what to do with it, what not to do, and what works?

One of the biggest levers: Claude Accelerators (this concept also works with other AIs like Copilot, ChatGPT or Gemini).

What I mean by this is to choose the 5-10 people on your team who already know what works, how, and where other tools are still needed (HTML, Python, Data Connectors, Cowork, Agents...).

They become your internal champions, the ones everyone else copies and the ones that can train your other teams!

I call them accelerators because they are the people that have the knowledge that we teach in the AI Finance Accelerator program that I run with Nicolas Boucher.

If you don't think you have these accelerators in your company, let me know and I'll send you more details of the course.

But if you already have them, then here's the plan you can adapt for your own company:

Week 0: Foundation​
Appoint an AI Finance Champion, recruit 5-10 Accelerators, procure Claude Enterprise. Bonus: get IT connecting Claude to your ERP or data warehouse early, it pays off from Week 4 on.

Week 1: Discovery​
Explore the Web App, Artifacts, Data Visualization, Connectors & Skills in a low-stakes way. Map where Claude is most powerful for your specific workflows.

Week 2: Build​
Ship one working MVP end-to-end: automated variance analysis + monthly review slides. Doesn't need to be perfect. Needs to run.

Week 3: Refine​
Polish that MVP to board-ready quality, document it so anyone can run it, and save the prompts as reusable Claude Skills.

Week 4: Scale Up​
Build a full driver-based forecasting workflow: base / upside / downside scenarios, automated narrative, live data if the connector's ready.

Week 5: Expand​
Automate three more core tasks: board pack, cash flow commentary, spend analysis. Add each to the Skills library.

Week 6: Culmination​
Reflect, quantify the wins, build your budgeting-season and year-end-close AI plans, and draft a 12-month roadmap tied to Q1–Q4.

Why does it work?

It's sequenced, not scattered. Each week compounds the last instead of throwing every feature at the team at once.

It follows the Pareto Principle. You chase the 20% of use cases that drive 80% of the value (variance analysis, forecasting, board packs) not every possible prompt.

It builds real deliverables, not just training. By Week 2 you have a working MVP. By Week 4, a forecasting workflow. By Week 6, a 12-month roadmap.

It creates champions, not just users.

I have a full set of materials that you can use for every week!

Here is the Excel version but feel free to email me back if you have questions or want specific materials on Skills, AI Agents, Python or anything else!

​Claude for Finance - 6 Weeks Adoption Plan by Christian Martinez.xlsx​

Thanks for reading,
Christian Martinez