If you've been named the AI champion at your accounting firm — or quietly handed the job without the title — this field guide is for you.
Jason Staats just published the six most common reasons accounting firms get stuck on AI. It’s a sharp list, drawn from more than a thousand accountants who came through his live events last year. And every blocker on it is written for the person who signs the cheque.
You’re probably not that person. You’re the one who got tapped on the shoulder — “you’re good with tech, can you figure out AI for us?” — and handed the problem with no budget, no authority, and not one hour taken off your existing work to go do it. From the owner’s chair, these are investment decisions. From yours, they’re something else entirely. That difference is the whole game.
The AI adoption blockers aren’t technical — they’re permission
Look at what that list becomes when you’re the one living it. “Where do I start” isn’t tool paralysis — it’s not knowing whether you’re even allowed to put client data into one of these tools. “We can’t find a champion” isn’t a vacancy — it’s you, the unicorn they went looking for, expected to produce magic with no mandate.
“The team won’t adopt it” includes you, for a rational reason: automate your own work and you just get handed more of it for the same pay, having made yourself look more replaceable in the process. “We need training” means you’re teaching yourself on your own time and quietly becoming everyone else’s trainer too. And “what’s left for my team” is really the question you lie awake on — what’s left for me?
Notice the pattern. Almost none of those are technical problems. You can learn the tools — what stops you is permission, time, recognition, and incentive, and those sit on someone else’s desk. That tells you exactly which problems to solve yourself and which ones to ask for help with.
Leading AI adoption starts where you don’t need permission
More is in your hands than it feels. Start in the gaps — the work that happens between your core platforms, on your own outputs, where no client data and no integration are involved. Drafting a client email, restructuring a messy file, turning a rough analysis into something a partner can read, summarizing a long document before a meeting. You don’t need a sign-off to point AI at your own draft work, and that’s where the fastest wins live.
When something works, don’t let it evaporate. Have the AI write up the steps so you can run it again — a reusable recipe the tools call a “skill” — and the second time costs you minutes instead of an afternoon. Then share it. That’s how scattered experiments turn into the visible wins the firm actually notices.
Don’t be the only battery, either. Jason’s right that one person carrying all the energy burns out fast. Find the one or two other people who light up about this and pull them in, and start a place — a channel, five minutes in a meeting you already have — where people name one thing they used AI for that week. You can build that habit without anyone’s budget approval.
And point the time you free up at the work a tool can’t do — the judgment calls, the client conversation, the read on a specific situation that only you have. That’s not a consolation prize. It’s the part of your job that gets more valuable as the routine work gets cheaper.
Then ask for the rest — and let the ask be the proof
The rest you ask for, plainly, as a short list. A green light on which tools are approved and one clear rule for client data, so you stop guessing. A few protected hours a week, with the mandate in writing. Recognition that counts at review time — and an explicit agreement that finishing faster won’t simply mean more work for the same pay. And a straight conversation about how your role grows from here.
Here’s the part to hold onto. The fact that you can write that list is proof you understand this better than anyone else in the firm. It isn’t a complaint — it’s the deliverable of the person who’s actually thought it through, and it’s a far stronger career move than quietly absorbing the whole problem until you burn out.
So if you’ve been doubting whether you’re “technical enough” to be the AI person, that was never the bar. The bar is whether you’ll do the part you control and ask clearly for the part you don’t.
And if you’re the partner reading over a team member’s shoulder right now — that list of four is what your champion won’t say out loud. Say yes before they stop asking.
To make that ask easier, I’ve turned the list of four into a one-page Ask List you can take straight to your partner — the exact asks, why each one matters, the words to use, and the “what I’ll do in return” half that makes it a fair trade instead of a wish list. You’ll find it with this article. Fill it in, and a hard conversation becomes a one-page proposal.
You didn’t need permission to read this far. What’s the first thing you’ll do without waiting for it?