DAVIDMERRY
ABOUT ME
Director of Finance and Chartered Accountant. I have been using AI every day for a few years - for traditional accounting tasks like reconciliations and automating month-end processes, and for more complex data analysis in Python and SQL that was not previously feasible.
WHAT'S HERE
Using AI in finance can be incredibly frustrating. There are influencers talking about how much of a game changer it is for financial models (it's always financial models), and the demos look fantastic.
When I started using AI, for this type of analysis, I got hardcoded numbers, no audit trail and formats all over the place. I spent as long patching it up as I would have done building it from scratch.
Soon after, I started treating AI as software I didn't know how to use, rather than software that didn't work. I spent the time learning it, watching videos, reading books, but mostly by trial and error (similar to how I learned Excel).
Now I think about whether a task actually needs AI before I start. If it does, I'm specific about:
- what the objective is
- what context (background information) it needs
- what the output should look like
What comes back is a lot closer to what I actually asked for.
That's what I write about here - what I've found AI to be good for, what it isn't, and my experience in getting the best out of it.
Mostly that comes back to using AI to build repeatable workflows that run each month, producing results I can verify.
I'd love to know what you think. If any of this is useful, or if you're doing similar work, reach out on LinkedIn.