I already knew that 3 people would be leaving.
We were 11 people in the finance team. And I already knew that 3 people would be leaving.
One of them was our chief accountant.
He knew so much about the ERP system and our accounting processes that everybody was worried about what would happen when he left.
"If he's gone, everything will really die."
People thought our processes would stop.
But three months after he left, we were still closing our books without a problem.
And eventually, we went from 11 people to 8, and then to 7. We were still doing the work, and we had more time to help the business.
How?
We challenged the work we should be doing.
We looked at what people were doing, what somebody else could do, what could be automated and – super important – what didn't need to be done anymore.
And I think this is a big part of what finance teams need to do as AI changes our work.
We need to challenge what humans should do, what machines could do and what should be done at all.
Sometimes the biggest improvement is to stop doing the work at all.
We were doing almost the same work twice
One of the biggest surprises was finding how much work was duplicated.
We had accounting on one side and controlling-FP&A, on the other. Both teams were doing their own processes. And they were not talking enough to each other.
So we brought them together. In fact, we put them in the same room. And we found that some processes were almost 80% the same!
Imagine that.
Two teams were carrying out almost the same process for different reasons. And because they were working separately, they were both doing the work.
Once we understood this, we stopped duplicating those processes. One team would do the work and inform the other about the result.
That was one of the changes that made a big difference.
And for me, it shows why you need to understand how work happens across a finance department. A person can be very good at their own process and still have no idea that the team sitting next to them is doing almost the same thing.
This is also why I don't think finance teams should become even more specialized.
When people only see their small part of the process, they don't see the whole picture. And that can also limit their judgment.
The solution is not always a new system.
Sometimes the people doing the work just need to understand what their colleagues are doing.
The work that didn’t need to exist
The second big surprise was how much we could stop doing.
For example, we were printing hundreds of pages and signing them. We also had reports and activities that continued because somebody had asked for them once or twice. And like this they became part of the normal process.
But that didn't mean we still needed them.
So we stopped doing some things. And for others, we reduced how often we did them.
Think about that for you.
You probably have processes that were created years ago. Maybe somebody needed a weekly report for a project. Maybe a manager wanted something printed. Maybe one person asked for an extra analysis.
Those requests can turn into routines that nobody questions. And when the team gets busy, you might start looking for ways to automate them.
But before you do that, you need to challenge whether the work is still needed. Ask who needs the output, why they need it and how often.
Of course, you still need to respect legal, audit and control requirements. But a habit is not the same as a requirement.
You don’t need to find a faster way to do every old task. Some tasks can disappear.
Today, AI gives us more ways to automate work. And this is amazing.
But you must question the work that needs doing first.
There's a simple process improvement framework that describes this. It's called ECRS:
- Eliminate work you don't need.
- Combine work that's duplicated.
- Rearrange who does what.
- Simplify the work that's left.
Only then should you ask what AI can automate.
A finance team needs people who understand more, not less
We also changed the responsibilities of people who stayed.
Before, some colleagues were focused on one part of accounting. For example, somebody might specialize in accounts receivable, while another person focused on accounts payable.
We started giving them more of the accounting function.
They could take on activities that they were already close to, without having to learn an entire process from zero. That made the team less dependent on any one person. If a colleague was on holiday, other people could cover the work.
And like this, people began to understand more of the accounting process instead of only knowing their own little part.
I think this is what the new finance team needs to become.
Instead of being super-specialized, finance people need to cover more.
Think of a finance pro as a T shape.
The vertical line of the T is your deep expertise in one main area, like accounts payable or controlling. The horizontal line is your working knowledge of the areas around it, such as accounting, FP&A and the business. You don't need to be an expert in all of them. You just need to understand enough to see how the processes connects, and to be able to apply your judgement.
You are still a specialist, but you are no longer stuck inside one small part of the work.
And with AI, it's now easier to develop those broader skills.
With AI helping you learn processes, ask questions and complete manual tasks, broader skills becomes much more possible.
And here's where I disagree with something I hear about junior finance people.
Some people worry that if AI takes over routine work, juniors won't learn.
I totally disagree with the idea that repeating the same process month after month is the best way to learn.
I remember learning a process after one or two months, then having to repeat it many more times. I wasn't learning much more from that repetition. I wanted to work on other things. I also remember how frustrating it was to use SAP when I didn't know the right transaction or which fields to enter. I had to keep asking somebody sitting next to me.
Today, AI could help me find that information and understand a process faster.
It could help juniors take on more subjects instead of waiting for somebody to explain every small task.
And that is a much better opportunity for learning.
But there is a risk.
Learning how to complete a task is not the same as learning how to judge the result.
When you look at an accounting number, you need to know whether it makes sense. You need to understand the business and know when to ask another question.
That judgment needs to be taught.
Managers and senior colleagues need to pass on their experience and critical thinking. Otherwise, you risk losing it as experienced people leave.
And let's not pretend the old way was perfect.
I've also seen people working in silos who were not developing enough critical thinking because they never saw the big picture.
So AI creates both an opportunity and a risk.
People could learn more, understand more and become better finance professionals.
Or they could complete more tasks without developing the judgment behind them.
The difference comes from how finance leaders train their teams.
The One Thing to Remember
Our team went from 11 people to 7.
That wasn't because we had generative AI replacing four people.
We changed the work.
We documented knowledge that had been held by one expert.
We brought accounting and FP&A closer together. And we stopped duplicate processes and activities that were no longer needed.
Then we gave people broader responsibilities so they could understand more of the business.
AI can now help finance teams make some of those changes more easily.
But it doesn't remove the need for leadership to decide which work matters, or to teach people how to think.
Before deciding what AI should do, challenge what your finance team should be doing at all.
Best,
Your AI Finance Expert
– Nicolas
P.S. The best way to learn from me directly is to join my weekly masterclasses. 60 mins + Q&A where I'll get all your AI in Finance questions answered. Join me here.
P.P.S. Want to see how a dashboard can answer your managers’ questions for you? Here is the video: I Added a CFO Assistant Inside Claude Dashboard
