A post called "Something Big is Happening" hit 80 million views last week. The author, Matt Shumer, is a tech CEO who builds AI products for a living. His claim: on a Monday morning, he described what he wanted built, walked away for four hours, and came back to find the work done. Not a rough draft. The finished product. Better than he'd have done himself. No corrections needed.
His conclusion: "I am no longer needed for the actual technical work of my job."
If you're running a CAS practice, your instinct is to file this under "tech world problems." A software CEO got automated by software tools. Interesting, not relevant. Except here's what happened inside your profession that same week.
Five products. One week. Zero humans required.
In the first two weeks of February, five companies launched autonomous bookkeeping products. Pilot unveiled a fully autonomous system — onboarding through monthly close, P&L, cash flow, balance sheet — with zero human intervention. Backed by Sequoia, Stripe, and Jeff Bezos. Digits published a benchmark showing their AI bookkeeper achieved 97.8% accuracy versus 79.1% for human outsourced accountants across 2,000 transactions. Canopy announced an AI bookkeeping module built into their practice management platform. BILL released agents that automate month-end close and eliminate over 80% of manual W-9 collection steps. Ramp launched its own AI bookkeeping solution.
The same week, Accrual launched with $75 million in funding to build AI-native tax preparation — agents that read K-1s, 1099s, and multi-hundred-page statements, identify what's missing, and produce draft returns ready for review. Goldman Sachs announced it's deploying Anthropic's Claude for accounting and compliance work, with their CIO saying he was "surprised" at how capable it was beyond coding.
That's not a forecast. That's a field report. And it maps precisely to what Shumer described in tech: the shift from "helpful tool" to "does the work better than I do" didn't happen gradually. It happened in a window that felt like weeks.
Your reference point is already outdated
Here's the part that should concern you most. If your last serious experience with AI in your practice was six months ago — or even three — you're making decisions based on a technology that no longer exists. The models available today are unrecognizable from what was available in mid-2025.
Shumer puts it bluntly: judging AI based on an experience from a year ago is like evaluating smartphones by using a flip phone. The Digits benchmark didn't exist six months ago. Pilot's autonomous monthly close didn't exist six months ago. BILL's W-9 agent didn't exist six months ago.
And here's the pattern that matters: these tools aren't perfect yet. Digits is at 97.8%, not 100%. Pilot still needs monitoring. The autonomous tax prep platforms need human review. But "not quite there yet" in AI has a very specific track record. If a tool shows a capability today, the next version will be significantly better at it. That 97.8% accuracy becomes 99.5% within months, not years. The trajectory only moves in one direction.
Every week your firm operates on assumptions formed from last year's AI capabilities is a week you're underpricing the speed of change.
The seen and the unseen
There's a useful framework from economist Frédéric Bastiat that applies here — the difference between what you can see and what you have to look harder to find.
The seen is obvious. AI compresses the commodity layer. Bookkeeping gets cheaper. Tax prep gets faster. Fee pressure on compliance work intensifies. Every CAS owner can picture this, and it's threatening.
The unseen is where the advantage lives. A Journal of Accountancy survey found that 79% of mid-market CFOs have already offloaded at least a quarter of their finance workload to agentic AI. But only 14% fully trust the output. That trust gap — between what AI produces and what someone is willing to stake their business on — is an advisory opportunity that didn't exist 12 months ago.
Someone has to validate the AI-generated financials before they go to a lender. Someone has to govern the agentic systems your clients are deploying without fully understanding. Someone has to be the wringable neck — the human whose name is on the work, whose judgment the client is actually paying for. That's you. That's always been you. But the scope of that role is about to expand dramatically, if you're positioned for it.
The firms staring at the seen — watching Pilot and Digits and panicking about bookkeeping revenue — are missing the explosion of advisory demand forming just outside their field of vision.
"Not quite there yet" has a shelf life
Shumer's post resonated with 80 million people because it named what a lot of professionals are feeling but haven't articulated: the ground shifted, and most people are still standing where they were.
The accounting profession is not immune. It was never going to be. The same AI capabilities that let a tech CEO walk away from his computer for four hours and come back to finished work are the capabilities that categorize transactions, reconcile accounts, draft client deliverables, and produce draft tax returns. Accounting isn't a separate problem. It's a downstream application of the same breakthrough.
But this isn't a doom story. It's a positioning story. The firms that learn to use these tools — not compete with them — and redirect capacity into advisory, governance, and the work that requires human judgment are building practices the rest of the profession can't touch. The ones that wait for the technology to "settle down" are going to discover that it settled on top of them.
If you don't have someone inside your firm leading this transition right now, that's the first problem to solve. Our AI Practice Transformation Program puts a trained AI Champion inside your practice in 15 working days — someone who can map your workflows, build the systems, and lead the shift from compliance revenue to advisory value. Fifteen days. One trained operator. A practice that's ready for what's already here. [Learn more at theaiaccountant.ai.]