Switch back to main site : Click Here
Every accountant and finance leader right now believes they should be doing something with AI. Most have no idea what that actually means for Tuesday morning.
That's the honest state of things. The pressure is real, the direction is vague, and underneath it is a quiet worry that's harder to say out loud: not should I automate? but am I already behind?
This piece is an argument that the anxiety is pointed at the wrong thing. The question that matters isn't whether AI is coming for accounting — it is, and it's going to be a genuine multiplier. The question is what makes AI actually work for you instead of past you. And the answer to that has very little to do with AI itself.
Start with what AI actually is in this context, because the hype makes it hard to see clearly.
AI is a power tool. And a power tool in the hands of someone with no plan builds nothing faster. In the hands of someone with a blueprint, it's a different job entirely. The tool is the same in both cases. What changes is whether there's something for it to work on.
A few things follow from that, and they're worth being plain about:
It won't fix a broken process — it'll run a broken process faster. Point AI at a chaotic, manual close and you don't get a modern close. You get chaos at scale. The mess has to be solved first. Then AI multiplies the solution. Eric Saunders, a CPA who runs his own firm (Balanced) on FinOptimal and uses AI more than most, puts it bluntly: "You can't just dump a GL into AI and say 'close the books.'" He's right — and he'd know, because he's built the thing that makes it work.
It doesn't replace judgment. It removes the work around judgment. The accountant's real value — the review, the exceptions, the genuinely new situation that needs a human to think — is the part AI can't do and shouldn't. What it takes off your plate is the rote, rules-based work that was never a good use of a human brain in the first place.
So the honest version is this: AI is inevitable, and it is a superpower. But it's inevitable and it's a genuine multiplier — only for teams with something for it to multiply. The infrastructure is what turns AI from noise into leverage.
That "only for" is the whole argument. Everything else in this piece follows from it.
Here's the part the AI conversation usually skips.
AI multiplies what you already have. Give it a continuous close — clean, current data and a standardized process — and it has something real to act on. Give it a traditional close — stale spreadsheets, workpapers rebuilt by hand, process living in one person's head — and it has nothing solid to stand on. Same AI. Radically different result.
This is why "garbage in, garbage out" becomes something sharper with AI in the mix: garbage in, garbage at scale. Layering AI on an unstandardized close doesn't produce a modern close — it automates the drift, the errors, and the key-person risk right along with everything else. You can't automate your way out of a process you never defined.
Which leads to the point that reframes the whole race: the infrastructure is the moat, not the AI. Everyone is going to have access to the same models. That's what a model is — a broadly available capability. What separates teams won't be whether they have AI. It'll be whether they built something worth pointing it at.
So the teams "waiting for AI to mature" have it backwards. The AI is ready. The real question is whether they are — whether they've done the unglamorous foundational work that turns a powerful model into actual leverage. For a firm, that foundation is standardized close infrastructure across the book. For an in-house team, it's a close that doesn't live or die with one controller's spreadsheets. Either way, it's the part that isn't automatic, and it's the part that takes real time to build.
This isn't hypothetical. Eric already runs it: his Stripe data flows from its source through a database, into a spreadsheet, and back into QBO as a single entry that updates daily — "every morning my clients see their month-to-date revenue booked in one neat summary entry without me lifting a finger." His words for it: "That's a continuous close. That's accounting in 2026." The foundation came first; the automation rides on top of it.
It's worth making this concrete, because "AI for accounting" is abstract to the point of meaninglessness right now.
Here's one real, near-term capability worth understanding: AI that flags inconsistencies and mistakes as the work is happening — not in a month-end review, but in the moment. A transaction categorized differently than it has been for the last eleven months. A number that lands outside its normal range. A duplicate that would have taken a human scanning hundreds of lines to catch. The system surfaces it, tells you why it flagged it, and recommends a fix — while there's still time and context to deal with it.
That's what AI on top of a solid foundation looks like. Not a robot doing the accounting. A second set of eyes that never gets tired, watching the books continuously, catching the things that used to surface only when something didn't tie at close. It works precisely because the foundation underneath it — current data, a standardized process — gives it a reliable baseline to measure against. On a chaotic close, "this looks inconsistent" is noise. On a continuous close, it's a save.
This is what Eric means when he describes the layer he's built on top of his own close: "If you explicitly instruct AI with skills, and then calibrate it to each client's nuances via memories, your team can cover more ground with more precision than ever." The AI isn't closing the books for him. It's carrying the repeatable work with a precision that only holds because the foundation underneath it is solid.
This is the direction FinOptimal is building toward. We built the continuous close first, on our own books, because it's the foundation. The AI layer on top of it — the part that watches, flags, and recommends — is what we're building for the future that's coming whether the profession is ready or not. More on that when it's ready to show. For now, the point is only this: the foundation is what makes any of it work.
If AI is a switch you flip once the foundation is built, the obvious question is: why not wait, and build the foundation later?
Because the foundation is the part with the lead time.
A real continuous close isn't an afternoon's work. It takes months to stand up — proving the process, standardizing across clients or entities, building the infrastructure that makes the data trustworthy. So "we'll get to AI later" actually means "we'll start the multi-month prerequisite later." The delay isn't on the AI. It's on the thing AI depends on. And that thing doesn't happen fast.
That lead time is why waiting costs more than it looks like it does:
The gap compounds. A team that starts now has a working process and clean data in six months — ready to compound from there. A team that starts in six months is where the first team is today, except the bar has moved in the meantime. This isn't a static finish line you can sprint to later. The distance grows while you wait.
The bar is rising underneath you. AI is already raising what clients and leadership expect — faster closes, more insight, lower cost. That's not a forecast; it's observable now. Meeting the standard you met last year is quietly turning into a losing position, because the standard didn't hold still.
"Same as last year" stops being the safe choice. This is the real shift. The status quo always felt safe because it was stable — doing what you did last month was the low-risk option. AI is what ends that stability. Same as last year used to cost you nothing. Now it's the most expensive thing on the menu; you just don't get the invoice until later.
None of this requires panic. It requires starting. Those are different things.
Come back to the worry we started with — the vague sense that you should be doing something about AI. Here's the resolution.
You don't have to figure out AI today. You have to start building the thing that makes AI matter, because that's the part with the lead time. And here's the good news buried in that: the foundation is worth building on its own terms, AI or not. A continuous close pays for itself in time back, trustworthy numbers, and a process that doesn't collapse when someone leaves — before you layer any intelligence on top of it.
So the sane path isn't a scramble. It's this: build the continuous close now. Get your data current and your process standardized. And when the AI layer is ready to do real work — watching the books, catching what's off, recommending the fix — you have something for it to amplify. AI becomes a switch you flip when you're ready, not a wave that catches you flat-footed.
The teams that win the next few years won't be the ones that found the best AI. They'll be the ones that were ready for it. The foundation is the continuous close. The intelligence goes on top. Start with the part that takes the longest, and the rest gets a lot less intimidating.
