I've been working in finance for almost a decade. For eight of those years, I've also been teaching finance professionals how to use data analytics, Python, and now AI to do their jobs better. Along the way, I've taught 200+ finance leaders, and the way I teach has changed almost as much as the tools themselves.

Here's what that journey has taught me.

Why Python Was Powerful but Hard to Adopt

Eight years ago, I started teaching finance professionals in Australia — sometimes in person, sometimes online — how to use Python for FP&A. I'd walk through all the benefits: automation, dashboards, faster and more consistent analysis.

Some people got it immediately. They started automating tasks, building dashboards, and running analyses they couldn't do before. But for many people, the feedback was consistent: it felt too technical. They didn't want to go that deep into code. They'd rather stick with Excel or spend their limited learning time on other skills entirely.

That feedback stuck with me because the underlying need never went away; the entry point was the problem.

How AI Changed the Entry Point

Three and a half years ago, ChatGPT 3.5 arrived, and I discovered it was genuinely good at writing Python code. That changed everything about how I taught.

I started teaching people how to use AI to generate Python code, which they could then use to automate workflows and run analysis, instead of writing or copy-pasting code themselves. People got excited in a way I hadn't seen before. Suddenly, it felt learnable because it was just about talking to AI and getting code back.

Here's the part that surprised me most: the substance of what I taught barely changed. I still taught people how to audit the code AI produced, how to make sure it connected properly to their systems, and how to ensure consistency. The only real difference was who wrote the first draft of the code: AI instead of them.

What 200+ Finance Leaders Taught Me

A few things have become clear from working closely with finance professionals as they picked this up:

1. Financial judgment does the heavy lifting, not the prompt. The people who get the most out of AI aren't the best prompters; they're the ones who already know what a correct forecast, a clean reconciliation, or a well-structured P&L should look like. That judgment is what catches AI's mistakes before they become someone else's problem.

2. AI-generated code still needs to be audited like any other code. Whether a human or an AI writes it, the same discipline applies: check the logic, check the assumptions, check that it connects properly to the source systems, and check that it produces the same answer twice.

3. The bottleneck moved from "can I write this" to "should I trust this." Writing the code got dramatically faster. Validating it did not. Most of the real skill-building now happens in that validation step, not in the coding step.

4. Seniority changes the outcome more than technical skill does. The professionals who go furthest with these tools tend to be the most experienced, not the most technical. Directors and fractional CFOs with strong financial instincts can direct AI toward the right questions and immediately spot when an output doesn't make sense.

5. Small automations compound quickly. People rarely start by automating something huge. They start with one recurring report or one manual reconciliation, and once they see it hold up over a few cycles, they extend the same approach across much larger parts of their workflow.

Where Financial Judgement Made the Difference

One participant, a fractional CFO working with a small manufacturing client, used AI to generate a Python script that pulled monthly sales data and built a rolling cash flow forecast, something that used to take her most of a day in spreadsheets.

The first version looked clean and the numbers seemed reasonable. But when she reviewed it against her own knowledge of the business, she noticed the model was treating a large seasonal customer order as a recurring monthly pattern rather than a one-off spike. Left unchecked, that assumption would have overstated cash availability for several months in a row, exactly the kind of error a business would only discover when the cash wasn't actually there.

She caught it because she understood the client's sales cycle, not because she reviewed the code line by line. Once she corrected the assumption and asked the AI to rebuild the logic around it, the forecast became something she trusted enough to bring into board conversations. The automation saved her time; her financial judgment is what made the output usable.

That's the pattern I’ve seen many times. AI compresses the time it takes to get a first version. It does not remove the need for someone who knows the business well enough to know when that first version is wrong.

What This Means for FP&A Leaders

The real capability here was never just learning Python, and it isn't just learning AI either. It's learning how to combine financial judgment, AI, and rigorous validation, so that speed doesn't come at the cost of accuracy.

Eight years of teaching this has convinced me that the tools will keep changing. What won't change is that the finance professionals who benefit most are the ones who treat AI as a fast first draft, not a final answer, and who never stop applying the same scrutiny they'd apply to a junior analyst's work.

If you'd like to go deeper on this topic, I write about it in my book, Smart Finance: The Origin — AI for Enhanced Financial Planning & Analysis, and teach it hands-on in my AI for Finance Accelerator program.