Accounting Just Got Its Copilot Moment
86% of tax professionals already use AI weekly. The shape change that hit software engineering just reached accounting.
86% of tax professionals are already using generative AI at least weekly. 36% use it multiple times a day. The US accounting workforce has shrunk 17% since 2020. The number of people sitting for the CPA exam is at its lowest point since 2006.
Read those four numbers together and you’re not looking at “AI coming for accounting.” You’re looking at the exact compression we just lived through in software engineering. AI adoption climbing fast, the human workforce shrinking faster, the work itself reshaping underneath everyone’s job title.
I’ve watched this happen twice now. Once in software engineering. And once last month, sitting at my desk doing taxes.
I Did Taxes for Three Businesses in Fifteen Minutes
Three businesses, plus my personal return. ChatGPT and Claude, nothing else. No tax software, no accountant doing the categorization, no spreadsheet hell on a Sunday. I exported transactions, dropped them into a chat, asked the right questions, watched the work happen.
That used to be weeks. Not because the work was complicated, but because it was boring and voluminous. Categorize the transactions. Reconcile against last year. Flag the deductibles. Match the schedules.
That layer is the accountant’s bottom rung. The same layer that built junior accounting careers for the last thirty years. And sitting there at fifteen minutes, I wasn’t surprised. I was recognizing something. I’d seen the same layer collapse in software engineering.
What Actually Changed in Engineering
When AI hit software engineering, nobody disappeared. The headcount didn’t crash. The job titles stayed the same. What changed was the shape of the work underneath the title.
Controlled studies put engineers with Copilot at 55% faster on coding tasks, on average a 1h11m task versus 2h41m without. That number wasn’t theoretical for long. The boilerplate collapsed first. CRUD endpoints, glue code, basic frontends, scaffold-and-wire projects, the stuff that used to fill a junior engineer’s week. Mostly gone now.
But the work didn’t disappear from the org chart. It moved up. Orchestration of agents running in parallel. System design across services that don’t yet exist. Verification, integration, the judgment the model can’t make. The seniors who absorbed the tools multiplied their output. The ones who fought them, or couldn’t get past “AI does my job,” got quietly squeezed.
That same fork is open in accounting right now.
What’s Collapsing and What’s Emerging
Watch what AI did and didn’t do during my fifteen minutes.
What it did: categorize, reconcile, flag deductibles, draft schedules, match prior-year patterns. The bottom layer of accounting. Collapsing.
What it didn’t do: tell me how to structure the businesses for next year, catch the multi-entity nuance where one company’s loss could shelter another’s gain, read the latest tax-code change and explain how my position shifted, take on an audit posture I’d actually trust. That’s the work emerging from underneath the collapse. The work that was always there, hidden beneath all the bookkeeping.
Proactive tax strategy across entities. Real-time advisory instead of quarterly catch-ups. Structuring conversations that used to be reserved for businesses doing tens of millions, now accessible to a five-person company because the busywork is free. Audit defense as a specialty. Fractional controllers for businesses that couldn’t afford one when bookkeeping ate the entire engagement budget.
The work didn’t shrink. It moved up the stack. Exactly the way it did for engineers.
Why This One Is Moving Faster
There’s a real difference from the engineering shift, though, and it’s the reason this disruption is moving faster than ours did.
When AI hit engineering, the workforce was growing. There was slack. Seniors absorbed the tools, juniors had a harder time getting in, but the system absorbed the shock.
Accounting doesn’t have slack. There are roughly 200,000 open accounting positions in the US right now. 75% of CPAs are nearing retirement. The exam pipeline is at its lowest point since 2006.
AI isn’t competing with accountants. It’s filling a gap that already exists. That’s why 86% of tax professionals already use it weekly — they don’t have a choice. The old shape can’t be held even if everyone wanted to hold it.
Who Builds the New Layer
If you’re an engineer reading this, you’ve already seen this movie. You know what happens next.
The bottom of the stack becomes commodity. The top of the stack multiplies. And in the middle, somebody builds the tools that connect the two, the new layer that didn’t exist before. In engineering, that “new layer” is now a generation of agent-orchestration products, AI-native dev tools, vertical Copilots. Companies founded in 2023 and 2024 that turned into the infrastructure of how software is built today.
The same wedge is wide open in accounting in 2026. Vertical AI tools for multi-entity tax strategy. AI-augmented controllership products. Real-time advisory platforms for SMBs that couldn’t afford one before. Audit-defense copilots.
This isn’t speculative anymore. Basis hit a $1.15B valuation in February for an agentic AI platform aimed at accountants. Accrual raised $75M to rebuild tax preparation as AI-native infrastructure. Pilot just launched its “AI Accountant,” a fully autonomous virtual worker. And Ian Crosby, the founder behind Bench, the last-generation bookkeeping startup that imploded, just raised $10M from Khosla to take a second swing with an AI-native version. The capital already sees the opening. The shape of what gets built next is still up for grabs.
If you’ve been close to accounting through a previous job, a family business, a startup you advised, that domain proximity just turned into a wedge. Same way years of healthcare or supply-chain experience turned into wedges when agents started moving past engineering.
The Real Question
Engineering was the first profession to walk through this fork. Accounting is the next, and it’s walking through faster.
There’s a third one. And a fourth. Law, design, consulting, financial planning, recruiting. Every knowledge-work profession built on a bottom layer of pattern-matching and form-filling is staring at the same shape change. Most of them haven’t noticed yet.
If you build software, the question isn’t whether to use AI in your job. You already do. The question is which profession you understand well enough, from family, from a previous career, from a side project, to build the new layer for, before everyone else figures out the door is open.
What’s the third one for you?
I’m putting together a side-by-side map of what’s collapsing and what’s emerging across knowledge-work professions, starting with engineering and accounting, then law, design, and consulting. If you want it when it’s ready, subscribe at macrostack.dev.