The question this raises for every professional services firm owner: are you integrated, or are you still piloting?

---

The Difference Between Piloting and Integrating

Most professional services firms that have been paying attention to AI are doing something. They have tools. They've tested workflows. They have a ChatGPT subscription or a Claude account or a specialized tool for their practice area.

That's not integration. That's piloting.

The distinction the Grant Thornton survey draws is between:

Piloting: AI tools are being tested in isolated use cases. Results are informal. AI use is optional and individual — some team members use it, some don't. The firm's core service delivery model is essentially the same as it was two years ago.

Integration: AI is embedded in core service delivery. The way the firm delivers value has been redesigned around what AI can now do. Specific workflows are formally structured around AI assistance. Outcomes are tracked. The answer to "how does AI affect how you deliver this service?" is concrete and specific.

The 4x revenue differential is between these two states. It is not between "uses no AI" and "uses some AI." A firm can have $5,000 in AI tool subscriptions and still be in pilot mode. A firm with one carefully integrated workflow can be further along the integration path than a firm with ten tools nobody uses systematically.

---

What Integration Looks Like at a 10-Person Firm

The Grant Thornton survey is based largely on enterprise organizations. "Full integration" at a 500-person consulting firm means something different than at a 10-person accounting practice. But the underlying logic translates directly.

For a 10-person professional services firm, integration is visible in three places:

1. The firm's standard service delivery has an AI component baked in — not optional

A 10-person accounting firm with integrated AI doesn't offer AI-enhanced bookkeeping as an optional add-on. AI-assisted bookkeeping review is how they do bookkeeping. The workflow is: client data comes in, AI processes the routine transactions and flags the exceptions, accountant reviews the exceptions and handles the advisory items. Every client, every engagement.

A 10-person law firm with integrated AI doesn't have one associate who runs contract review through an AI tool when they feel like it. Contract review includes an AI pass as a standard step before attorney review. Every contract, every time.

2. Roles have changed — and the firm has made something of the change

When AI handles the deterministic, routine work in a professional services firm, the hours that work consumed become available for something else. Integrated firms have made a deliberate decision about what that "something else" is. More clients at the same head count. Deeper advisory work per client. New service offerings that weren't feasible when all the capacity was consumed by routine work.

Firms still in pilot mode haven't answered this question. AI saved some time somewhere, but that time hasn't been redirected into anything systematic.

3. Client outcomes have measurably improved and the firm can articulate how

An integrated firm can tell you: since integrating AI into our tax prep workflow, we turn around preliminary returns 8 days faster, and our clients receive their first-pass return review before we request additional documents, which reduces the back-and-forth by an average of two rounds. They can tell specific clients specific things about how their engagement improved.

A piloting firm can tell you: "We've been experimenting with AI and it's been helpful." That's not integration.

---

The Governance Gap That Could Undo the Revenue Upside

The survey's second major finding is less discussed and more urgent for professional services firms specifically.

78% of business leaders said they lack confidence they could pass an independent AI governance audit in 90 days. Nearly 75% have granted agentic AI access to their data and processes. Only 20% have tested their incident response plan if an agent fails.

For professional services firms, this gap has a specific liability dimension that enterprise technology companies don't face in the same way.

When an AI agent operating in a professional services context — a law firm, an accounting practice, a consulting firm — produces an incorrect output that a professional reviews and forwards to a client, who owns the error? Professional licensing standards, malpractice frameworks, and client agreements all say the licensed professional owns the output. That's true regardless of whether AI produced it.

A professional services firm that has given agentic AI access to client data without a documented failure protocol has not just a governance gap — it has an unassessed professional liability exposure.

The 20% incident response test number is the most actionable finding in the survey for professional services firm owners. Not because AI agents are unreliable — they aren't, within their designed scope — but because failure protocols are what demonstrate that a firm has thought through professional responsibility in its AI deployment. They are the documentation that shows a firm exercised appropriate oversight.

The 20-minute incident response baseline:

1. For each AI tool that handles client data or produces client-facing output, identify what "failure" looks like — incorrect output, unexpected behavior, data handling issue
2. Write down how you would detect that failure before the output reached a client
3. Write down what you would do if the failure was detected after the output reached a client
4. Name who is responsible for each step

That's not an enterprise risk management exercise. It's a 20-minute conversation with whoever manages your AI tool stack, documented in a shared folder. It puts your firm in the 20% — ahead of the 80% that haven't done it.

---

The "Wait and See" Conversation Is Over

The single most useful thing the Grant Thornton survey data does is close the "wait and see" argument.

"Wait and see" was always a resource allocation argument: the cost of integrating AI now might exceed the benefit if the technology changes, if the tools don't work for our specific use case, if competitors don't move either. For small professional services firms, it was a reasonable position in 2023. It was defensible in early 2024.

The 4x revenue growth differential, measured across 950 organizations in early 2026, shows what happens at the population level to firms that kept waiting while others moved to integration.

Integration doesn't require a technology project or a consultant. For a 10-person professional services firm, it requires three decisions:

1. Which one service do you deliver that has the most automatable routine work?
2. What would it look like to rebuild delivery of that service with AI handling the routine work as a standard step?
3. What would you do with the capacity that frees up?

The data says firms that make those decisions and follow through grow at four times the rate of firms that don't. The case for deciding is about as clear as survey data gets.

---

What to Do This Week

If you're an accounting firm: Identify one recurring service — bookkeeping review, tax preparation, payroll reconciliation — where AI could handle the routine transaction work as a standard step. Spend one hour this week building that workflow on paper: what does AI handle, what does the accountant review, what happens to the time saved? That's the integration decision. Everything after that is implementation.

If you're a law firm: Identify one document-heavy service — contract review, due diligence, discovery — where AI could complete a first pass as standard practice. Most small law firms already have a tool that can do this. The question is whether it's optional or standard. Make it standard for one matter type this month.

If you're a consulting or staffing firm: Apply the three-question integration diagnostic: (1) What work in my firm is deterministic enough for AI to handle reliably? (2) What have I done with the capacity AI has already freed up? (3) What client outcomes can I point to that improved because of AI? If you can't answer all three specifically, you're still piloting.

The survey closes the debate. The firms that integrate outgrow the firms that pilot, by a factor of four. The question is how soon you want to be on the right side of that number.

---

Related: Where Does Your Firm Fit on the AI Adoption Curve? The Thomson Reuters 2026 Benchmark | How to Track AI ROI in a Professional Services Firm | KPMG Just Cut Its Auditor's Fee 14% Using AI as the Argument