Two Researchers From MIT and Stanford Measured What AI Does for Small Accounting Firms. Here's What They Found.

Published: September 13, 2026 | By: The Crossing Report

Most AI research you'll read about accounting comes from vendors. Vendors have a product to sell, which means their data has a direction. The numbers are often real — they just aren't randomly selected.

That's what makes this study different.

Two researchers from MIT and Stanford published independent findings in the Journal of Accountancy in August 2025. No vendor sponsored it. No software company's product is named as the winner. They wanted to know what happens — in actual firm data, not a controlled demo — when small accounting firms adopt AI tools.

Here's what they found.

---

The Study: 79 Firms, 277 Accountants, Real Transaction Data

A field study from MIT and Stanford researchers analyzed transaction-level data from 79 small-to-midsize accounting firms and survey responses from 277 accountants.

The firms studied range from 5 to 50 employees — the same range as most professional services firms served by The Crossing Report. These aren't enterprise clients with innovation departments and dedicated AI teams. They're firms like yours: partners managing client relationships, staff doing the work, and owners trying to figure out whether the AI investment is worth it.

The researchers tracked actual firm performance before and after AI adoption. Not self-reported satisfaction scores. Not demo output. Ledger data, close timelines, and client load per accountant.

Three major findings came out of that analysis.

---

Finding 1: AI Cut Monthly Close Time by 7.5 Days

AI adoption was associated with a 7.5-day reduction in monthly close time.

If your monthly close currently runs three weeks, that number means you could be done in less than two. If it runs ten days, you could compress it to under three.

That's not a rounding error. That's a different operating model.

The reduction is driven by the parts of close that take the most time and require the least professional judgment: data entry, bank reconciliation, transaction categorization, and draft preparation for review. These tasks are where accountants spend enormous amounts of time doing work that is important but not intellectually demanding. AI handles the pattern recognition, the categorization, the matching — and flags exceptions for human review.

The implication for your firm runs in two directions:

Option 1: Same number of clients, faster turnaround. A faster close means you can deliver monthly reporting to clients sooner, which improves the client relationship and differentiates you from competitors still running on a three-week timeline.

Option 2: Same timeline, more clients. If you recover 7.5 days of close capacity per month, that capacity can be redirected. Which brings us to the second finding.

---

Finding 2: Accountants Supported 55% More Clients Per Week

Accountants supported 55% more clients per week after AI adoption.

This is the number that should change how you think about growth. Most firm owners think about adding clients in terms of adding staff. The math is simple and wrong: one more senior accountant costs you $80,000–$120,000 per year, onboarding takes three months, and you're back to capacity constraints within a year.

The firms in this study didn't grow by adding staff. They grew by recovering time that was being consumed by manual tasks.

A 55% increase in client capacity means a firm currently serving 20 clients per accountant could move toward 31 — without hiring. For a 5-person firm where each accountant handles 15 clients, that's potentially 8 additional clients per person, 40 additional clients firm-wide. At $2,000/month average revenue per client, that's $80,000 in monthly recurring revenue from existing headcount.

The math is not guaranteed. Results depend on what your current time is actually spent on, which workflows you automate first, and how consistently your team adopts the tools. But the directional evidence is strong: AI-assisted accounting firms are serving more clients with the same people.

---

Finding 3: Accountants Shifted 8.5–9% of Their Time Toward Higher-Value Work

AI shifted 8.5–9% of accountant time from routine data entry toward higher-value advisory work.

Paired with that shift: ledger granularity improved by 12%. Finer, more accurate financial records came out of the same workflow — not because accountants were more careful, but because AI-assisted categorization caught distinctions that manual entry tends to blur when volume is high.

These two findings are related. When accountants spend less time on data entry, they have more time to review what the data means and explain it to clients. That's the shift from compliance to advisory — not as a strategic aspiration, but as a measurable output of automation.

The 12% improvement in ledger granularity also has a compounding effect: more accurate records make advisory conversations more credible and more useful. When you can show a client a clear trend in their cost structure that previously got lost in broad categorizations, you're providing value they couldn't get from the old process.

This is the part of the AI transition that firms consistently underestimate. The efficiency gains are obvious. The advisory leverage is less visible but ultimately more valuable — and it's documented in this data.

---

What This Means for Your Firm This Quarter

Three workflow changes produced most of these results. None of them require a full firm technology overhaul.

1. Automate bank reconciliation first. This is the highest-volume, lowest-judgment task in most close processes. AI-assisted reconciliation — available in most modern accounting platforms — matches transactions, flags exceptions, and eliminates the manual matching that consumes hours per client per month. Start here. Measure time before and after for 30 days.

2. Use AI for transaction categorization review, not entry. Rather than having accountants categorize each transaction manually, use AI to categorize at scale and have accountants review the output for exceptions. The review step maintains accuracy and professional judgment while eliminating the entry bottleneck. This requires training the AI on your firm's categorization conventions — typically one to two weeks of setup per client.

3. Move the advisory conversation earlier in the close cycle. When close takes three weeks, the advisory conversation happens at the end — as an afterthought, after the client has already been waiting for numbers. When AI compresses close to one week, you can deliver preliminary findings in the first week and use the remaining time for analysis and conversation, not data work. This is the structural shift that enables the advisory revenue expansion the study documents.

The 7.5-day reduction, the 55% client capacity increase, the shift toward advisory work — these aren't separate phenomena. They come from the same change: removing manual data tasks from accountants' time and redirecting that time toward work that requires professional judgment.

The firms in the MIT/Stanford study that achieved these results started with one of these three workflows. They didn't redesign their practice. They automated one thing, measured it, and expanded from there.

That's the model. This quarter, pick the one workflow that consumes the most time per client in your close process. Pilot AI-assisted tools for 30 days. Measure the difference. Then decide what comes next.

---

Sources: CFO Dive, "AI cuts monthly financial close time by 7.5 days: MIT/Stanford study" (cfodive.com); Wisconsin Institute of CPAs summary of the Journal of Accountancy research; MIT Sloan Ideas Made to Matter. Study published Journal of Accountancy, August 2025.