It’s hard to find balanced takes on what AI means for finance teams in the short term.
Over on LinkedIn, people are fully losing their minds.
My feed is filled with fantasies of CFOs firing their whole team after a heavy weekend of vibe coding.
Let’s be clear: AI is going to transform the CFO's office. But not in the ways you’re reading about. Not yet.
Like every tech wave before it, Amara’s Law will apply. We overestimate the short-term productivity impact and underestimate the long-term one.
AI is right on cue. And this time, the curve may well be steeper. But that doesn’t make today’s hype any more helpful.
You’re being told to fire half your team. Rebuild your stack from scratch. Learn to code like an engineer. For CFOs that is nonsense.
Very little AI tech for CFOs is truly enterprise-ready yet.
If the world’s best engineers - backed by billions in funding and full-stack AI teams - are still building, what makes you think you can crack it with two mugs of coffee and a 45-minute YouTube Replit tutorial?
The idea that a CFO or their finance team should be building proprietary AI software isn’t just unrealistic... today it’s borderline reckless. You’re not a product org. You’re not a startup. You’re the steward of your company’s finances. Your job is to protect the business’s assets and grow shareholder value.
To be clear, this advice is for established CFOs and senior finance leaders. If you’re early in your career, or you’re a current or aspiring CFOTech founder? You should absolutely be experimenting more aggressively. That’s a different playbook (and not one I should write.)
But if you’re already in the seat. Running a finance team, managing cost pressure, tariff chaos, audit reform, and a volatile boardroom... your job is not to vibe-code your way through it.
Now let’s talk about the “95% myth.”
There’s this idea floating around that AI can do 95% of a finance task in five minutes - board packs, forecast models, strategy decks, you name it.
But in finance, precision beats direction. Just because something looks 95% done doesn’t mean it is. Worse, it gives a false sense of confidence. You think you’re nearly there. You’re not.
The final 5% is where all the risk lives. All the value. All the judgment. That’s the bit that matters most.
CFOs aren’t paid for the first 95%. That’s the commodity layer. They’re paid for the last 5%, the sharp end of the decision. The edge.
And the best-paid CFOs, are paid seven or even eight-figures ones for the final 0.1%. The part that only experience and real-world judgment can deliver.
So if you shouldn’t be hacking your team to death with AI experiments, what should you be doing?
Here’s the practical AI playbook for CFOs today as I see it:
1. Reframe where AI can actually help. It can’t write your board paper. But it can help you tighten it. It won’t build your model. But it can sanity check it. It won’t find your narrative. But it can pressure test it.
2. Don’t trust the “95%” myth. People say AI gets you 95% of the way there. But that’s an illusion. Just because something looks 95% done doesn’t mean it is. And in finance, that last 5% - the nuance, the judgment, the risk - is everything. It’s not the final touch. It’s the whole game.
3. Let the builders build. Take the demos. Work with tech vendors who live and breathe finance problems. Your job isn’t to build an AI-powered P2P system from scratch. It’s to pick the best one on the market and implement it well. We are a generation of CFOs scarred by horrendous tech implementations. But AI will make implementation and customization infinitely easier than we've seen before.
4. Get your data organized. AI loves structured data. The more your data is clean, consistent, and well-structured, in a proper data warehouse layer, the more AI-ready you actually are. Think integrity, granularity, metadata. That's what will unlock what comes next.
5. Experiment in low-stakes environments. Play where failure doesn't hurt. In my case, I built custom AI bots to help my kids with homework, tailored to their curriculum, learning preference, instructed to guide to the answer not give them the answer, etc. That’s stretched my thinking and showed me the art of the possible, more than any finance workflow. It’s where I’ve learned the most. You want cheap feedback loops, fast iterations, low risk.
And for what it’s worth, here’s where I’ve seen AI be genuinely useful so far:
- Editing comms: emails, specs, board packs
- Sorting and prioritizing inboxes
- Stress-testing story outlines and presentations
- Surfacing directional insights from messy data
- Scoring candidates against hiring criteria
- Acting as a thinking partner when I need to challenge my own assumptions.
So no, don’t fire your FP&A team and vibe-code a workflow replacement. Experiment, yes. But pick simple workflows. Low stakes. Clear feedback loops.
Don’t eat the hyperbole. Keep one eye on AI, not both eyes and both hands. Your job is to run the business, not chase the unproven LinkedIn fever dreams.
Keep your hands on the wheel.
Stay curious. Stay skeptical. But most of all ... stay very f*cking crispy.