CFOs have been asking me this same question for 3 years.

"How do I upskill my team on AI without creating chaos or security risks?"

Over the last few years, I've trained finance teams at Mercedes-Benz, KPMG, Rakuten, AWS, and hundreds of other companies on AI implementation.

Just last month, I helped 4 companies take their teams from AI beginners to advanced users with this 30-day program.

By the end, they'd built use cases that surprised even me!

An invoice analytics dashboard, an automation to book journal entries and an agent that reviews legal contracts.

But, what I'm STILL seeing today is that most finance leaders have no idea where to start.
​

​
Excitement ≠ ROI

The trend in 90% of companies is where leadership gets excited about AI. Then, they buy a load of licenses, send an email announcing the new tool… and then…

Nothing happens.

Sure, your team members that are super interested in tech, they figure this out on their own. But, everyone else doesn't know where to start - or - doesn't have time to experiment - or - doesn’t see the value.

So they quietly ignore the whole thing. Which means that the licenses sit there unused.

This is crazy! And then leadership wonders why they are not seeing any ROI from their spend.

This pattern is everywhere, and it's completely avoidable.

It is more than possible to see big improvements in way that finance work.

Providing you follow a structure that works.

Here’s your 4 week blueprint.

​

​
Step Zero: Pick One Tool and Commit

Before anything else, you need to decide which tool your team will use and give everyone a license.

You have two paths here. If your company is a Microsoft or a Google Workplace user, use what you already have (Copilot or Gemini).

The security is already approved, you don't need to bring in a new vendor. Plus, the AI has access to your emails, calendar, and files. This is really valuable, because AI can help with your work rather than just random tasks.

If you’ve not gone into Microsoft Copilot or Google Gemini and asked “Search my recent e-mails for anything I have missed”, you can do this now and thank me later.

Only if you need to procure something separate (for specific use cases that Copilot or Gemini cannot do), then go with ChatGPT Business/Enterprise or Claude Enterprise.

Only these versions have SOC 2 Type II security standards, which is the level you need to use AI professionally with sensitive data.

I wrote a full breakdown on choosing the right tool here in my previously newsletter “Which 1 AI tool should your finance team use? (hint: you already have it)”

The key is to pick one and commit. Don't let your team confuse themselves with five different tools.
​

​
Week One: The Eye Opener

You cannot give your team a license and expect them to figure everything out by themselves.

The first thing you need to do is organize what I call the Eye Opener session. This is a one-hour session where you show people what AI can do beyond writing emails.

I always start with a "banger" demonstration using my six-step approach for financial analysis, because when people see AI generate a dashboard from their data in minutes, people REALLY want to be able to do this themselves.

What to cover in the session:

• Upload data and ask AI to check data integrity
• Ask AI to propose analysis strategies
• Run calculations and generate commentary
• Create a visual dashboard from the results
• Teach a basic prompting framework people can follow

Don't skip this step, because if you just hand out licenses without showing people what's possible, most of them will never get started.
​

​
Week Two: Hands-On Workshop

The Eye Opener is the first 10% where you share what's possible. But we all know that we only learn by doing, so now you need people to try it themselves.

Run a hands-on workshop where people apply what you showed them in week one, but this time using their own files and solving their own problems. This is where the real learning happens, because people will have questions and run into roadblocks, and when you troubleshoot together they will learn much faster.

How to structure it:

• Create 3-4 exercises based on real tasks from your company
• Give people 15-20 minutes per exercise
• Troubleshoot together as a group
• Let people learn from each other's questions

Your team is busy with their day jobs, and most of them will not find two hours on their own to experiment with a new tool. You need to organize that time for them.
​

​
Week Three: Build the Sharing Habit

By now you've created some momentum. Your people have licenses, and they've seen what's possible, plus they've tried it themselves.

But this progress will die if you don’t maintain it (and keep people interacting).

Launch a Slack or Teams channel where people share their best use cases and ask questions. In any team you'll have people who pick things up quickly and people who need more time, and you can't be the manager standing behind everyone. You need to let people learn as a group.

Rules you can announce:

• When someone shows you something cool, tell them to post it in the channel
• When someone asks you a question directly, redirect them to ask publicly
• If you answer privately, one person learns. If you answer in the channel, everyone learns.

We have an area for this in the AI Finance Club, it’s a dedicated area called “Share your use cases”.

Here’re are a few examples of some recent wins:

Someone in senior finance used AppScripts to save a tonne of time during month-end. See Monday’s newsletter here if you missed how to do this.

Plus, a CFO automated their morning routine using AI after reading this newsletter at the beginning of this year.

I also recommend adding a recurring live element to keep the channel from going quiet.

My favorite is a monthly lunch and learn, where you gather your team, order pizza (or people bring their own food), and have informal chat for the first half. Then someone presents a use case or a new feature for the second half - or - you can even just watch a tutorial together and discuss how to implement it.

​

​
Week Four: The Hackathon

Let's finish with something fun. A hackathon where everyone builds a Custom GPT (or Copilot Agent, or Gemini Gem).

This is the most underused feature in all of these tools. I've trained a lot of people and even those who use AI every day almost never use it. But it’s one of the most powerful things you can do.

Think about it this way. You don't start Excel reports from a blank spreadsheet every time. You start from a template. But most people start every AI conversation from zero, even when they're doing the exact same task they did yesterday. A custom GPT is like starting with a template where you just change a few inputs to get the output you need.

How to run it:

• Give your team two hours and a framework
• Everyone builds a mini agent for a task they do repeatedly
• Vote on which ones are the best
• Make the winners available to the whole team

Three things happen from this.

1. Everyone learns how to build these tools.
2. Some people will create genuinely useful things you can start using immediately.
3. The team will realize that, if they could build something this good in two hours, maybe there's an opportunity to invest more time and make these tools available to everyone.

To give you an idea of what's possible, here's what people built in a single hackathon I ran at Autodoc:

• A dashboard analyzing 64k lines of vendor invoices, flagging PO vs No PO
• A Tax Review Gem with a built-in checklist and risk assessment (what used to take 2-3 hours now takes 30 minutes)
• A KYC process that previously took 3-5 days
• Automated accruals booking by vendor
• An IFRS 16 review tool that translates and summarizes lease contracts

All of this, built by finance pros (not engineers) in just a few hours.
​

​
The One Thing to Remember

You don't need the biggest budget or the most technical talent.

You don't need a 6-month transformation plan from a consulting company charging you $100k to steal your watch and tell you the time.

All you need is structure, and accountability.

You can't just hand out licenses and hope people figure it out themselves, and you can't run one training session and expect habits to form overnight.

30 days and 6 steps will take you from a team that's heard of AI to a team that uses it every day.

I want finance teams to be proud of themselves. I want other departments to look at what you've built and feel super jealous, wondering how your team got so far ahead.

That can be your team. But only if you start.

If you're passionate about training and have the bandwidth, you now have everything you need to run this yourself.

But if you'd rather have a pro (like me) organize this for you, reach out or reply: hello@nicolasboucher.online​

​

Best,

Your AI Finance Expert,

Nicolas

​

P.S. - What did you think of this? Will you use it? Hit reply (I read all replies)

P.P.S. - If you're worried you're falling behind with AI, and want to stay ahead in 2026. Join my Masterclass 'How to Become an AI CFO' here.