Most of the conversation around AI in accounting has focused on the wrong question. "Will AI automate bookkeeping?" is now settled. Yes, it will. In many firms, it already has. And for the firms that haven't adopted it yet, their clients are starting to, which answers the question from a different direction entirely.
The more useful question is what happens after. The core work that made accounting firms essential for decades (pulling data, tying out numbers, reconciling accounts, closing the books) is rapidly becoming something most accounting software can do. That's not a threat to the profession. It's a threat to a specific version of the profession, the one where the prep work consumed so much of the calendar that it became the product.
What clients have always wanted is someone who can tell them what the numbers mean. Can they afford to hire, is their growth sustainable, and does their entity structure still make sense. They want judgment, accountability, and someone who understands the full picture. That work isn't going away. As AI makes baseline financial services cheaper and faster, more businesses will have clean books for the first time and need someone to help them understand what's in them. Demand for real accounting expertise is about to expand, not contract.
The firms that capture that demand will be the ones that made three decisions early.
Where your intelligence lives
There's a meaningful difference between a firm that uses AI tools and a firm whose core system actually supports AI. Most accounting platforms in use today were architected before AI was a consideration. They store transactions as rows of text with no native ability to understand what a transaction represents, who the counterparty is, or how it relates to the client's broader financial picture. You can add AI features on top of that, and many vendors are, but the underlying system can't learn, can't reason across clients, and can't improve over time.
The platform you run on determines the ceiling for every AI capability you'll have access to. A firm evaluating technology right now shouldn't be comparing feature lists. They should be asking whether the system's data model supports intelligence natively, or whether every smart feature is a workaround bolted onto a foundation that wasn't designed for it.
How you price the work
This is the decision most firms are avoiding. If AI turns a ten-hour engagement into thirty minutes of review, hourly billing collapses. Fixed-fee models aren't safe either, not if the fee is just a proxy for estimated hours. Clients are watching AI transform every other service they use, customer support, software development, legal work, and so on. They will eventually ask why they're paying the same monthly rate when the labor underneath has fundamentally changed.
The firms in the strongest position will be the ones pricing based on outcomes: clean books, accurate financials, timely filings, strategic insight. Not the labor, but the result. That's harder to implement, but it's the only model that stays coherent as automation accelerates. Firms that set these terms now, while they're choosing to, will be in a very different position than firms that get dragged there by client pressure later.
The same logic applies to technology vendors. A platform that charges a subscription regardless of whether it reduces your workload has no structural incentive to make your firm more efficient. A platform whose revenue depends on delivering measurable automation does.
What happens when your best people leave
The accounting profession has lost over 300,000 practitioners since 2020. Graduates are at a 20-year low. The average age of a CPA firm partner is north of 55. The implication most firms haven't acted on: every year, institutional knowledge is walking out the door with no mechanism to retain it.
The judgment calls a senior partner makes, the quality standards a manager enforces, the industry-specific conventions a specialist carries: if those only exist in people's heads, they depreciate with every departure. The firms solving this are encoding their expertise into the technology that does the work, so the system applies the firm's standards automatically, across every client. That turns institutional knowledge from a liability that shrinks over time into an asset that compounds.
The bigger picture
These three decisions reinforce each other. A modern architecture enables higher automation. Higher automation justifies outcome-based pricing. Outcome-based pricing creates the incentive to encode expertise into the system. Encoded expertise makes the automation smarter. The whole thing compounds.
There will be meaningfully fewer accounting firms in the US within the next few years. Some will transform. Some will scale by acquiring others. Some will exit. PE firms are already using technology stack as a valuation metric. The gap between technology-forward firms and everyone else is widening, not closing.
The firms in the best position won't be the ones that adopted the most AI features. They'll be the ones that made the structural decisions about their ledger, their pricing, and their knowledge that enabled them to deliver the work clients actually hired them for.
The opportunity is genuinely enormous. But it belongs to the firms that move, not the ones that wait for certainty before starting.
