6 Lessons from my own business
Two weeks ago I did something I don't usually do.
I got on a call with more than 80 people from my AI Finance Club, and before I even started sharing my screen, I told them: this is confidential, don't share the recording, some of this you won't even find inside the club later. And then I opened up use cases from my own business.
Why would I do that? Because I'm a finance guy first, and finance guys don't trust a system until they've seen how everything works. So I showed them how it works for my business.
Since then, I decided I needed to share with you too.
I'm not going to give you all 55 slides. But I picked six things from that session that I think will add real value to your finance function.
Let's go.
1. Build your digital twin
|
|
Here's the thing I told everyone straight away: if you take one action item out of today, it's this one.
Somewhere in your business, there's a project, a GPT, a Copilot agent, a Claude skill that knows everything about how you run things. It might not even be in AI, it could be in a Google Doc or OneDrive.
I have mine. I call it my digital twin, and I use it constantly, for things like strategy, contracts, creating org chart infographics etc.
Now think about your own function for a second. How much of "how we decide things here" lives only in your head? That's concentration risk, the same than if 80% of your revenue sat with one customer. A digital twin won’t replace your judgment – it makes your judgment queryable by you and the team, helping you save a tonne of time and energy.
One rule, though: build it once, then improve it every month. Keep updating and adding context, and your twin will become more intelligent over time.
2. Decide upfront where AI can and can’t work
|
|
We get around 500 emails a day. We have a workflow tool (n8n) that reads and classifies every single one of them before a human ever sees it.
But we decided in advance what AI is allowed to answer on its own. Somebody wants a fast reply about membership? AI will answer this. Somebody wants corporate training for their team? A human will answer this.
This is exactly the discipline you already run for spending approval – a delegation of authority matrix. Under this amount, a manager signs. Above it, it escalates. Most finance functions don't have the same for AI. They're either blocking it everywhere out of fear, or letting it answer everything because nobody wrote down the rules.
3. Build mini-consultants
|
|
My wife is a consultant, and she helped me build something I actually think is one of the most useful things in my whole business: a "problem-solving" project. Inside, it has every root-cause framework, plus step-by-step instructions on how to walk through one.
Somebody on my team, with zero experience in structured problem-solving, used it recently to actually find the root cause of an access issue we were having (and fixed it). Using AI for your team could be as simple as giving them a framework to act on.
For you you could build mini-consultants for:
- Walking the team through a process for spotting anomalies.
- Running month-end efficiently.
- How to build Excel models according to your business rules.
4. Trade the 10 page memo for a 1 page infographic
|
|
Somebody on my team was explaining me, on Slack, how one of our internal systems stores information. They kept typing. I kept not understanding. So I asked: can you just make me an infographic that shows how it works?
With one image, I understood straight away. Same thing happened with a huge strategy document we'd built with AI for the future of the club. It was too long for anyone to read start to finish, so instead we turned it into one visual and used that in every conversation since.
If you're the one presenting a model, a plan, or a case for change to your board or your exec team, a lot of people will nod along because the memo was 12 pages and nobody wants to admit they skimmed it?
A one-page infographic is a super fast way to get a concept across, and it also reduces the risk of misunderstanding.
5. Move your best people from doing to reviewing
|
|
We had two files that were supposed to show the same member count. They didn't. I asked someone on my team (who has never done an audit in their life) to reconcile them with AI. Something that would have taken them, and me, ages to do by hand.
They came back with a clean file. I reviewed it. And it was done.
The hardest part of all of this is your best people stop being the ones doing the reconciliation, the variance analysis, the sentiment tally across hundreds of comments. AI does the first pass. Your people (and you) become the review layer.
Really important to know that is not a downgrade of the role. That is literally the control function auditors have used for decades, you are just moving it earlier in the process.
But – and I always say this – the review has to be real. Not a quick look and a sign-off. If you skip the review because "AI did it," you haven't reduced any risk, you’ve just added another step.
6. Schedule the adoption (and be honest about what doesn’t work)
|
|
We run a lot of dashboards, designed by people who are not developers. And sometimes they don’t work. Sometimes a number looks wrong, and you go in and find out it's been wrong since the beginning.
I'm not telling you this to scare you. I'm telling you because the alternative (pretending it's all perfect) is worse. My team knows things break. And we talk about them openly.
What makes adoption happen aren’t Slack updates. It's protected time.
We have exactly one recurring meeting across the whole team: one hour, every week, which is our AI learning hour. This is the only recurring meeting for the team. That's it.
If you want your finance function to actually use this stuff in six months, don't wait for enthusiasm to spread on its own. Put one hour on the calendar, every week, and don't cancel it.
The One Thing to Remember
The best thing you can do when trying to adopt AI in Finance is share.
Hackathons to show AI uses cases, weekly AI learning meetings, sharing wins in Slack and Teams channels. These will all help.
As you all know, I’m a numbers guy that prefers to keep things super practical.
And I was unsure whether to share all of this.
But I hope that by doing this it has inspired you, and shows that the only way to move from theoretical to practical is to start building, and share.
Best
Your AI Finance Expert
– Nicolas
P.S. – This week my community crossed 3,500 members, and my daughter drew me a picture to celebrate!
Join me here next Tuesday for my free masterclass, and I will explain at the end how you can join.
|
|
P.P.S – Ready to start building? Here's the full Claude guide for finance → Watch This Claude Guide For Finance If You Want to Be Ahead of 99% of People