Scaling accounting capacity with OpenAI

Accounting isn't exactly the sexiest corner of the AI boom, but it might be one of the most practical. Basis, a startup building on top of OpenAI's model stack, just published results showing its AI agents are saving accounting firms as much as 30% of their working hours. That's not a marginal efficiency gain — that's roughly a third of a workweek freed up, per employee, at firms that are notoriously understaffed and drowning in repetitive reconciliation and compliance work.

The technical detail here matters more than it might seem. Basis didn't just bolt a chatbot onto spreadsheets. It stitched together four different OpenAI models — o3, o3-Pro, GPT-4.1, and the newer GPT-5 — presumably routing tasks based on complexity and cost. Reasoning-heavy models like o3-Pro likely handle the gnarly judgment calls (tax nuance, anomaly detection in ledgers), while GPT-4.1 or lighter GPT-5 configurations probably chew through high-volume, lower-stakes drafting and categorization. That kind of multi-model orchestration is quietly becoming the norm for serious enterprise AI products, rather than the single-model demos that got most of the attention in 2023 and 2024.

What's notable is the framing OpenAI and Basis are pushing: this isn't about replacing accountants, it's about capacity. Accounting firms have spent years turning away clients or delaying advisory work because staff are buried in compliance busywork. If agents can genuinely absorb 30% of that load, firms get to redirect senior staff toward the stuff that actually grows revenue — advisory, planning, strategic consulting — instead of hiring more junior accountants just to keep pace with tax season.

There's an obvious tension baked into this pitch, though. Every AI vendor selling into professional services says some version of "we free you up for higher-value work," and sometimes that's true, and sometimes it's a polite way of saying headcount growth slows down. Accounting firms operate on thin margins and billable hours, so a 30% time reduction is going to ripple into staffing decisions whether anyone frames it as "growth" or not. OpenAI publishing this as a customer story signals it wants enterprise buyers, not just consumer hype, to be the next growth engine — and professional services, with its mountains of structured, rules-based work, is fertile ground for that pitch.