Most AI-finance pitch decks have a version of the same slide: "~50,000 CPA practices in the US, plus tens of thousands of solo accountants" (iPassTheCPAExam), rolled up into a "$200B accounting services market," sold via "per-seat SaaS to firms."
The trouble is that "accounting firm" is not one buyer. Bench burned through capital selling cash-basis bookkeeping to solopreneurs and shut down in 2024 (Steph's Books); Pilot moved toward venture-backed startups precisely because the unit economics of "small-business owner" and "Series B CFO" don't match (Sacra). Tools sold generically as "practice management for accounting firms" (Canopy) address firms spanning a 1000x range in revenue per partner, with different billing models and different decision-makers. This piece separates the universe into the three buyers the economics actually produce.
Disclosure: I'm building Artifi, an AI tooling layer for finance teams, so I have a stake in this market. The observations below come from selling into firms across all three profiles.
Profile A: independent small firms
The biggest segment by count. 81% of AICPA MAP Survey respondents in 2025 were firms at or below $5M in revenue (AICPA). Most bill per client: a flat monthly fee for bookkeeping, an annual fee for the report, hourly for one-offs.
That billing model creates a structural tension with automation. When revenue is bounded by client count rather than accountant productivity, making accountants more productive doesn't grow revenue — it exposes overstaffing, unless the firm can raise prices (hard for small practices) or add clients (which needs sales capacity these firms rarely have). I engaged three Estonian firms in this profile over the past year; each hit a different blocker — positioning, technical literacy, and the billing-model trap — and none could adopt. The detailed cases are the subject of a separate piece on the innovator's dilemma. The pattern that matters here: in Profile A, the buyer's economics often penalize the efficiency the software provides.
Profile B: big advisory firms
A different shape entirely. The top 100 US firms generated over $100B combined in 2025 (Accounting Today), Deloitte alone around $33B (Inside Public Accounting). These firms bill on value, carry a fast-growing advisory mix, and can monetize automation as more billable advisory hours rather than fewer accountants.
They are also building at industrial scale. EY reports 1,000+ AI agents in production with plans to scale far higher and over $1B/year invested; PwC, KPMG, and Deloitte have their own platforms (ChatFin summary). For a vendor, the Big 4 are less a customer than a benchmark. The more reachable buyer is the band beneath — Top 25 to Top 200 firms. In one engagement with a Top-25 US firm, the objection was not staffing economics but integration surface area: whether the APIs would survive a security review and connect to NetSuite, Sage Intacct, and custom warehouses. The deal logic is standard enterprise SaaS — long cycle, legal and security review, custom connectors, partner buy-in — gated by integration, not by the firm's economics.
Profile C: PE-backed rollups
The newest shape, and the one fewest vendors target explicitly. Private equity has been consolidating accounting firms rapidly — the deal data and mechanics are covered in detail in PE-backed accounting rollups. What matters for buyer segmentation is that the rollup inverts Profile A's blockers:
- Profile A wouldn't automate because accountants would leave; the holdco treats reduced labor intensity as part of the plan.
- Profile A wouldn't share its UI because the firm wants to be the interface; the holdco rebuilds the interface once across many portfolio firms.
- Profile A's staff couldn't adopt the mental model; the holdco hires a centralized team that can.
The deal shape differs accordingly: typically one decision-maker (a holdco CPTO or COO), one platform decision across N firms, an ACV that justifies real implementation, and a timeline measured in weeks because operational leverage is the investment thesis. It is also where the land-and-expand logic is strongest — a single workflow adopted at the holdco can widen across every firm in the portfolio.
Different GTM, pricing, and product
The three profiles tend to require different motions. Profile A: direct sales rarely pencils out for either side; where it's touched at all, it's usually through the white-label tooling of the platforms these firms already use, at low per-seat prices with self-serve onboarding. Profile B: enterprise sales on a 6–12 month cycle, landed via an innovation or partnership champion, priced on ACV with integration services, with a security review measured in quarters. Profile C: founder-led sales into the holdco, a shorter cycle when the value prop is clear, and platform-tier pricing structured so the holdco benefits as it acquires more firms — which assumes a multi-tenant management layer built early.
The product surface differs too: Profile A leans on an embedded API, Profile B on an enterprise console with role-based access and SSO, Profile C on a tenant-of-tenants management plane with consolidated reporting. A single "one thing for everyone" build tends to fit none of them cleanly.
What this means for each group
For vendors: the data suggests pipeline is worth tagging by profile, because A, B, and C convert through different motions at very different ACVs; treating them as one funnel tends to mask why deals stall.
For independent firm partners: the structural options that recur in this segment are consolidation (roll up or get rolled up) or repositioning around fixed-fee advisory — both covered more fully in the innovator's dilemma piece.
For investors: the fastest-growing capital base in the category sits in Profile C, while most vendor pipelines still skew to Profile A — a mismatch worth probing in diligence.
The takeaway
The single "accounting firm" buyer in the TAM slide doesn't really exist; the economics produce three. They differ on who decides, how value is billed, what integration is required, and whether automation helps or threatens the model. The distinction is less a marketing nicety than the thing that determines which motion works — and which segment quietly consumes sales effort without converting.