76% of Small Businesses Use AI. Only 14% Have Actually Integrated It. Here's the Difference.

There's a number buried in a Goldman Sachs survey from early 2026 that every professional services firm owner should see: 76% of small businesses use AI. Only 14% have fully integrated it into core operations.

That gap — 76% to 14% — is not a motivation gap. It's not a budget gap. It's not even a technology gap. It's a structural gap. Most small firms are using AI the same way they use a search engine: one person, one task, one result. The 14% are doing something fundamentally different. They've changed how they deliver services.

This post explains what that difference looks like in practice, why the three barriers the survey identifies are symptoms rather than causes, and the one decision that moves a firm from the 86% to the 14%.

---

What the Goldman Sachs Survey Actually Found

The data comes from the Goldman Sachs 10,000 Small Businesses Voices survey, conducted January 27 through February 4, 2026, by Babson College and David Binder Research. The sample was 1,256 participants from the Goldman Sachs 10,000 Small Businesses cohort — owners who had gone through the program's business education curriculum, meaning these are experienced, serious operators, not hobbyist side-hustle runners.

The headline finding: 76% of these small businesses currently use AI tools. That number is high — and it's genuinely encouraging. Small business owners are not sitting this out.

But here's the number that matters: only 14% have fully integrated AI into core operations.

That's not a coincidence. It's a pattern. And the survey data explains why.

The positive signal most people miss

Before getting into the gap, it's worth noting what the survey also found: 93% of small businesses using AI say it had a positive impact. 84% cite increased efficiency and productivity as the primary benefit. 67% expect AI to increase their revenue.

This is important. The story is not "AI doesn't work." The story is "AI is working for almost everyone who uses it — but something is preventing most of them from capturing the full value."

That something is integration.

The training gap: 73% want more help

The survey found that 73% of small business owners say more training and resources would help them implement AI more successfully. That's not a complaint. That's an opportunity.

It also points directly at the structural problem. If nearly three-quarters of small business AI users feel under-equipped to implement it, the barrier isn't access to tools. The tools are there. The barrier is knowing what to do with them — specifically, how to move from "my staff uses AI occasionally" to "AI is part of how we deliver this service."

The three named barriers

The survey identified three primary barriers to full integration:

1. Data privacy and security concerns (cited by 50% of respondents)
2. Lack of technical expertise (49%)
3. Difficulty choosing the right AI tools (48%)

All three are real. But none of them is the root cause. More on that below.

---

What "Fully Integrated" Actually Means for a Small Firm

The term "integration" gets used loosely. Let's be precise about what it means — and doesn't mean — at a 5-to-30-person professional services firm.

It does not mean: AI installed on computers. Staff with ChatGPT subscriptions. A few prompts saved in a shared Google Doc. An AI tool mentioned in your marketing materials.

Integration means: AI is part of how you deliver a specific service. It's not a personal productivity tool that individuals choose to use or not use. It's embedded in the workflow — with a defined process, a quality standard, and a person responsible for it.

Here's a practical test. A firm with integrated AI can answer these three questions:

1. Which service line is AI now part of standard delivery for? (Not "we use AI for various things" — a specific, named service.)
2. What is the workflow? (Not "my team uses AI to draft things" — the actual sequence of steps, inputs, outputs.)
3. Who is responsible for the quality of that AI-assisted work? (A named role or person, not "everyone.")

If you can answer all three for even one service in your firm, you're more integrated than 86% of your peers.

The revenue evidence

This isn't a theoretical distinction. Grant Thornton's 2026 AI Impact Survey — conducted independently of Goldman Sachs, earlier in 2026 — found that firms with fully integrated AI are 4x more likely to report revenue growth compared to firms still in the piloting phase (58% vs. 15%).

The mechanism makes sense. When AI is integrated into delivery, the firm can take on more work with the same team, turn it around faster, and maintain consistent quality — all of which creates revenue capacity that a firm where "employees use AI sometimes" doesn't have.

---

The Three Barriers Are Real — But They're Not the Problem

Let's go back to the three barriers the Goldman Sachs survey named, because each one looks different once you understand the underlying structural issue.

Data privacy and security (50%)

This is a legitimate concern. Professional services firms handle client confidences. Uploading sensitive financial records or case files to an AI tool without understanding where that data goes is a real risk.

But the 14% solved it. They didn't solve it by hiring a security consultant or waiting for a perfect solution. They solved it by designing workflows where client data stays within firm-controlled systems — using tools that don't train on inputs, setting up enterprise agreements with clear data handling terms, or structuring AI into the parts of the workflow that don't touch sensitive client data (research, drafting from firm-approved templates, internal analysis).

The privacy barrier is surmountable. It requires a deliberate decision, not a technology breakthrough.

Lack of technical expertise (49%)

Nearly half of small business AI users say they lack the technical know-how to implement AI successfully. This feels like an insurmountable problem until you look at what the 14% actually did.

They didn't hire engineers. They didn't build custom AI systems. They picked one tool, learned it deeply, and applied it to one service. The firms that solve the expertise barrier aren't technically sophisticated across the board — they're deeply competent in a narrow application.

The firms that stay stuck on "lack of expertise" are typically trying to evaluate AI broadly — "what should we use AI for?" — rather than narrowly — "how do we use Claude/ChatGPT/Copilot specifically for this one thing we already do?"

Breadth is what kills technical confidence. Narrow focus builds it.

Difficulty choosing the right AI tools (48%)

This is the most revealing barrier, because it describes a trap.

Tool selection paralysis is not a technology problem. It's a strategy problem in disguise. The firms that have been stuck in tool evaluation mode for 12 months are not stuck because the tools are unclear — there are now well-established, capable tools for almost every professional services workflow. They're stuck because they haven't committed to a specific use case, so every tool seems potentially useful and no tool seems definitively right.

The 14% stopped evaluating tools. They started running workflows. They picked something — not because it was objectively the best option, but because it was good enough for the specific thing they wanted to do — and they built the workflow around it. Once the workflow exists, the tool evaluation question answers itself.

Tool paralysis is a symptom of not having made the delivery decision yet.

---

The One Decision That Changes Everything

Every firm that has crossed from the 86% to the 14% made the same decision, even if they didn't name it that way.

They picked one service line and committed to rebuilding its delivery model around AI.

Not "we'll explore AI across the firm." Not "everyone should try AI in their work." One service. A real redesign of how that service gets delivered. A named workflow. A named owner.

This is the distinction between a tool decision and a delivery decision.

Most firms make tool decisions. They buy a tool, assign it to staff, and wait for results. The tool generates some time savings and useful outputs, but nothing about the firm's fundamental service delivery changes. The tool is an add-on.

The 14% made delivery decisions. They looked at a specific service — maybe it's monthly financial close reporting for accounting clients, maybe it's lease agreement review for a small real estate law firm, maybe it's market research for a consulting client engagement — and they asked: "If we were designing this service from scratch today, knowing what AI can do, how would we deliver it?" Then they rebuilt it.

That's the crossing moment. When AI moves from being a personal productivity tool to being delivery infrastructure.

How to identify your integration-ready service line

Not every service line is equally ready for AI integration. Here's a quick diagnostic:

A service line is integration-ready if:

• It involves a defined, repeatable process (not entirely bespoke judgment each time)
• It produces a documentable output (a report, an analysis, a document, a recommendation)
• The quality of that output can be assessed by a trained professional
• You deliver it multiple times per month

For a 10-person accounting firm, that might be cash flow projections for advisory clients, or tax research memos, or new client onboarding documentation. For a consulting firm, it might be competitive landscape analyses or executive briefings. For a staffing firm, it might be candidate shortlisting summaries or client outreach sequences.

What to do in the first 30 days

Once you've named the service line:

1. Document the current workflow — every step, every input, every output, who touches it and when
2. Identify where AI replaces or augments steps — specifically, not "AI helps with writing," but "step 4 (drafting the analysis section) now uses Claude with this prompt template, reviewed by X before delivery"
3. Run it 10 times — enough to refine the process and identify the failure points
4. Name the owner — one person is accountable for the quality of AI-assisted work in this workflow

That's it. Not a company-wide rollout. Not a policy manual. One service, documented, tested 10 times, one owner.

---

What This Means Heading Into H2 2026

The Goldman Sachs survey was fielded in January and February 2026. The data reflects where small businesses stood at the start of the year. We are now past mid-year.

The gap is not shrinking. It's widening.

The Federal Reserve's April 2026 monitoring note found that 40.43% of professional services firms report AI usage — suggesting even the usage number is lower in professional services than in small business broadly. The firms at the top of the adoption gap — the 14% with real integration — are now 12+ months ahead of the firms still in pilot mode.

Here's what 12 months of integration head start means: they've built the workflows, trained the staff, worked out the quality controls, and started redeploying the capacity that AI freed up. They've had time to use that capacity to take on more clients, launch a new service line, or raise their fees while improving delivery margins.

The Grant Thornton data — firms with fully integrated AI are 4x more likely to report revenue growth — describes compounding. The gap that exists today will be larger in 12 months.

But here's the other thing the data says: 86% of your competitors are in the same place you are. The window hasn't closed. The firms that make the delivery decision this quarter join a very small cohort that is currently pulling away from the field.

That is the crossing. Not from "not using AI" to "using AI" — 76% of small businesses have already made that crossing. The crossing that matters now is from "using AI" to "running on AI." From tool to infrastructure. From individual productivity to service delivery redesign.

That's the one you haven't made yet. And right now, 86% of your competitors haven't made it either.

---

Start Here

Pick one service. Name it. Write down the current workflow. Identify where AI enters it. Run it 10 times. Name the owner.

That's not a six-month project. That's a 30-day decision.

The data says the firms that make it are 4x more likely to grow revenue. The data says 86% of your competitors haven't made it yet.

The Crossing Report covers exactly this transition every week — specific workflows, real implementations, and the tools that professional services firms are using to make it. Subscribe here to get weekly intelligence on how firms like yours are crossing from using AI to running on it.

---

Source: Goldman Sachs 10,000 Small Businesses Voices survey, conducted January 27–February 4, 2026 by Babson College and David Binder Research (n=1,256). Covered by CPA Practice Advisor, Benzinga, and Fortune (March 2026).

Related Reading

• The AI Adoption Gap in Professional Services
• AI ROI for Professional Services Firms