The Federal Reserve Mapped Small Business AI Adoption — And 93% of Firms Are Still Experimenting

Published: May 2026 | By: The Crossing Report

---

The Federal Reserve Bank of San Francisco just handed professional services firm owners the most credible benchmark they're ever going to get — and the number that matters most isn't the one getting the headlines.

Yes, 46% of small businesses now use AI. That's the stat everyone will quote. But here's the one worth writing down: only 7% have fully integrated it. That means 93% of small firms using AI are still experimenting. They're using a tool here, running a pilot there, maybe saving a few hours a week — but they haven't crossed the line from "AI as a productivity add-on" to "AI as a core part of how we run the business."

That gap — between using AI and integrating it — is the most important number in the SF Fed's March 2026 report "Early Findings on Small Business Use of AI." And it maps directly to the most common place professional services firm owners get stuck.

This piece explains what the data actually shows, what full integration looks like for a 15-person professional services firm, and how to move from the 46% to the 7%.

---

The Headline Number: 46% of Small Firms Use AI — But Only 7% Have Crossed the Line

The SF Fed data comes from the 2026 Federal Reserve Small Business Credit Survey, collected from thousands of small business owners across the US. Unlike vendor surveys, this is neutral government research — the kind of data that gets cited in congressional testimony, not used to sell software.

The core findings on AI adoption:

• 46% of small firms use AI in some capacity
• Only 7% have fully integrated AI into their business operations
• 71% of small firms using AI report productivity gains
• An additional 15% of non-using firms plan to adopt AI within 12 months

Here's what those numbers tell you: the majority of small business owners who are using AI are somewhere in the middle. They're experimenting. They've seen the productivity gains — 71% report them, which is a remarkable number. But they haven't built AI into the core of how they work.

The 46% figure will get cited as evidence that AI adoption is accelerating. That's true. But the 7% figure is the one that deserves your attention. It tells you where the competitive advantage actually lives, and how far most firms — including most of your competitors — still are from reaching it.

What "Fully Integrated" Actually Means for a 15-Person Professional Services Firm

The Federal Reserve's framework distinguishes three stages:

Experimentation: One or two people use an AI tool informally. There's no policy, no training, no standard workflow. The tool is being tested, often by the most tech-curious person on the team.

Partial integration: AI is embedded in specific workflows but hasn't touched core operations. A consulting firm's researchers use AI for background briefs. An accounting firm's staff uses it for email drafts. The tool saves time, but intake, client delivery, and billing still run the old way.

Full integration: AI is embedded in core business processes — intake, service delivery, client communication, and financial operations. It's not a tool that some staff use. It's part of how the firm runs. A 15-person accounting firm at this level is using AI to accelerate onboarding, draft deliverables, flag review items, and produce client-ready summaries without adding headcount.

The 7% who've reached full integration aren't using better tools than the 39% in partial integration. They made a different organizational decision: they stopped treating AI as an add-on and started treating it as infrastructure.

---

Why the Gap Between Adoption and Integration Is the Real Story

The 39-percentage-point gap between "uses AI" (46%) and "fully integrated AI" (7%) isn't a technology problem. It's a change management problem.

The General Assembly 2026 survey of professional services firms found that 61% abandoned at least one AI project in the past year. The #1 reason wasn't cost, technical complexity, or lack of tools. It was internal change management and stakeholder communication.

That finding should resonate with any firm owner who has tried to get their team to consistently use a new tool. You can pick the right AI platform, pay for the subscription, and do the first-week training — and still find six months later that three people use it daily, two use it occasionally, and the rest have quietly reverted to the old way.

The Three Stages Most Small Firms Are Stuck Between

Most small firms that use AI get stuck at one of three transition points:

From nothing to experimenting: The barrier here is usually the first decision — which tool, for what workflow. This is where most of the "I know we should be using AI but..." conversations happen. The fix is a forced constraint: pick one tool, one workflow, 30 days. Not a strategy. A test.

From experimenting to partial integration: This is where the tool gets expanded beyond the early adopter to the broader team. The barrier is usually training and buy-in — making a tool part of how everyone works, not just the curious few. This requires explicit workflow documentation and a standard process, not just "here's a subscription, figure it out."

From partial to full integration: This is the hardest transition because it requires rethinking core workflows, not just adding AI to existing ones. A firm making this crossing typically has one person (often the owner) who decides that the old intake process, the old delivery workflow, or the old billing review needs to be rebuilt around AI — not enhanced by it. This is the transition most firms never make.

The 7% who've crossed into full integration made it through all three.

---

What 71% Productivity Gains Look Like Without Full Integration

The SF Fed's finding that 71% of AI-using small firms report productivity gains is real — and it applies at every stage of integration, including experimentation.

That's the good news. You don't have to reach full integration to benefit from AI. Partial integration delivers measurable results.

The Upside Firms Are Leaving on the Table

Here's the nuance: partial integration captures the easy productivity gains — the hours saved drafting, researching, and summarizing. Full integration captures the compounding gains — the ones that change what the firm can actually offer clients.

A 10-person accounting firm in partial integration might save 4–6 hours per week across the team on email drafts, meeting notes, and research summaries. That's real. It adds up to meaningful capacity over a year.

The same firm in full integration — where AI is embedded in client onboarding, tax preparation review, and advisory delivery — doesn't just save hours. It can take on more clients without adding staff, turn around deliverables faster, and offer services (AI-assisted advisory, proactive alerts, continuous monitoring) that it couldn't offer at all before. The difference isn't incremental. It changes the firm's growth ceiling.

This is what the data means when it shows 7% at full integration reporting disproportionately higher productivity gains than the 39% at partial integration. The upside isn't in the tool. It's in how deeply you've built the workflow around it.

For a framework on tracking what that productivity actually translates to in dollars, the AI ROI measurement guide for professional services firms covers the specific metrics worth tracking.

---

How Professional Services Firms Move From Experimenting to Integrated

The path from 46% to 7% is not about adding more tools. It's about going deeper with the ones you have.

The One-Workflow Starting Point That Unlocks Full Integration

If your firm is in the experimenting or partial-integration stage, the fastest path forward is picking one workflow that touches every client and rebuilding it around AI.

For a professional services firm, that workflow is usually one of three:

Client intake and onboarding. Most firms have a manual intake process — a questionnaire, a phone call, a kickoff meeting, followed by a lot of data entry. An AI-integrated intake workflow uses a structured intake form that feeds directly into AI-generated client context documents, reducing the time from "signed contract" to "first deliverable" by days.

Deliverable drafting and review. Whatever your firm produces — tax returns, audit reports, consulting deliverables, legal documents — there's a first-draft stage that can be AI-assisted and a review stage that can be AI-flagged. Full integration means AI generates a first draft from the client data and flags items that need human attention, rather than a human starting from a blank page and AI being used occasionally.

Client communication and reporting. Regular updates, status emails, and periodic reports are often the most time-consuming part of client service delivery. An AI-integrated communication workflow uses templates and context documents to draft these automatically, leaving the human to review and personalize — not to write from scratch.

Pick one of these three. Rebuild the full workflow — intake to output — around AI. Document it. Train the team on it. Use it on every client for 60 days. That's the transition from partial to full integration in one workflow, which is the foundation for everything after.

The barriers most firms hit when adopting AI — specifically the data security concerns and the client trust hesitation — are real but solvable. The Karbon data on how accounting firms navigate those two specific fears is worth reading before you rebuild a client-facing workflow.

---

What the San Francisco Fed Is Watching Next (May 19, 2026)

The SF Fed's March 2026 research isn't a one-time report. The San Francisco Federal Reserve is hosting a public event on May 19, 2026: "Small Business and AI: Understanding Adoption and Building Support Systems."

The framing of that event title tells you something important. "Building support systems" is not about the technology. It's about the organizational and infrastructure support that small businesses need to make the crossing from experimenting to integrated. The SF Fed's research conclusion is the same as what firm owners report on the ground: the technology is available. The gap is in the support infrastructure — training, trusted vendor guidance, implementation help, and clear workflow models.

That a Federal Reserve bank is convening a public event specifically on small business AI support tells you how seriously government economic researchers now take the adoption-integration gap. This is not a technology story anymore. It is a small business competitiveness and economic development story.

The sources you should bookmark: the SF Fed's full March 2026 article and the Federal Reserve's Small Business Credit Survey chartbooks are the primary data. These are worth saving — they'll be the citation behind every credible small business AI conversation for the next 18 months.

---

The Crossing Report Verdict

The SF Fed data is the clearest version of the core argument: most small firms using AI are not using it in a way that changes their competitive position. They're using it the way a firm owner might use a calculator — helpful, but not transformative.

The 7% who have fully integrated it are running a different kind of firm. They're not necessarily smarter or better resourced. They made a decision to go deeper — to rebuild a workflow instead of just adding a tool to it.

The 93% of AI-using firms still experimenting is not a failure. It's an open door.

What to do this week:

Identify the one workflow at your firm that every client touches. Write down every step in that workflow from intake to delivery. Mark every step where a human is starting from a blank page, transferring data manually, or drafting from memory. Those are your AI integration points.

That list is your roadmap from 46% to 7%.

You don't need a consultant, a technology committee, or a six-month plan. You need a single workflow rebuilt around a tool you're probably already using. Start there.

---

The Crossing Report is a weekly intelligence newsletter for professional services firm owners navigating the AI transition. Free subscribers get the top three insights each week. Premium subscribers get the full analysis, implementation guides, and tool comparisons — including the AI workflow rebuild playbook.

Subscribe to The Crossing Report →

---

Related Reading

• AI ROI Measurement for Professional Services Firms
• How Professional Services Firms Are Building AI-Native Business Models
• AI Staff Adoption in Professional Services