What the 20% Capturing AI's Value Are Actually Doing

74% of the economic value from AI is flowing to 20% of companies. The other 80% aren't sitting on the sidelines – they're using AI too.

That's from PwC's 2026 AI Performance Study (PwC 2026 AI Performance Study, PwC, April 2026), and it's the part of the AI conversation that hasn't gotten enough attention. The lagging firms aren't underinvesting in tools. They're running the same tools differently – layered on top of their existing workflows, existing delivery models, existing pricing structures. And that's where the returns collapse.

The architecture problem

When thinking.inc analyzed how professional services firms are deploying AI, they found a return range of 40 to 350 percent depending on one variable: whether firms changed their pricing model alongside the deployment. (AI in Professional Services, thinking.inc, 2026)

That's not a technology finding. It's a commercial architecture finding – and it tracks exactly with what Grant Thornton concluded in their 2026 analysis: "The proof gap isn't a technology problem. It's a commercial architecture problem." (Grant Thornton 2026 AI Impact Survey, Grant Thornton LLP, 2026)

Most advisory practitioners run their businesses around time. Hours billed, retainer blocks, project fees calibrated to effort estimates. AI compresses the time required for a lot of that work without changing the value delivered. If your pricing model stays effort-based and your workflow stays sequential, you've made work faster without capturing any of the upside. The tool changed. The architecture didn't.

The 40-to-350 percent spread isn't random variance. It's what happens when the same technology meets two different commercial structures.

What the 20% actually changed

PwC found that top performers were twice as likely to redesign their workflows and delivery models around AI rather than simply adding AI to existing processes. (PwC 2026 AI Performance Study)

That's the whole distinction, compressed into a single data point. Not which AI tools. Not how much they spent. Whether they changed how they work.

In boutique advisory specifically, what that redesign usually involves is the extraction layer – the part of an engagement where you're reading source materials, building landscapes, cross-referencing documents, structuring raw data before the actual strategy work begins. In most complex engagements, this phase is the heaviest lift. It's real analytical work. It requires intelligence and domain familiarity. What it doesn't require is the specific combination of senior judgment, relationship history, and strategic interpretation that the fee is actually earned on.

The firms capturing real AI ROI figured out how to handle the extraction layer differently – with AI as the synthesis engine, not as a faster way to do what they were already doing. The judgment work is still theirs. The intelligence layer feeding it changed.

The boutique constraint

Here's what makes this harder for boutique practitioners than for large enterprises: the advantage is real, but so is the constraint.

Boutique advisory firms have things large enterprises can't buy – faster decisions, niche expertise, deep client relationships, the ability to act on new intelligence within days rather than quarters. thinking.inc's analysis of professional services performance confirms that these structural advantages compound when they're backed by the right knowledge infrastructure. (AI in Professional Services, thinking.inc, 2026)

But building that infrastructure properly takes more time and dedicated resources than most boutique shops carry. Enterprise firms absorb the investment because they have dedicated teams and capital budgets. A 3-to-5-person boutique – or a solo practitioner – doesn't have the runway to build internal AI tooling from scratch.

The practitioners who've navigated this aren't building their own knowledge management systems from scratch. They're partnering with synthesis capabilities that already have the infrastructure in place, the domain familiarity, the AI tooling – and applying those capabilities to their specific engagement context and client history.

Practices with purpose-built knowledge infrastructure compete on different terms. Each engagement deepens the context layer – the patterns from one client's work inform the speed and calibration of the next.

What this means in practice

Grant Thornton's framing holds: if the returns on AI aren't materializing, the first question isn't "are we using the right tools?" It's whether you changed anything about how you deliver work, price your services, or structure your intelligence layer when you started using AI.

Most boutique practitioners haven't. Not because they're slow or resistant to change – but because layering AI onto an existing workflow is genuinely easier than redesigning around it, and because the ROI data that would motivate the harder work has mostly been published about large enterprises.

Knowledge infrastructure for a boutique advisory practice isn't a technology stack. It's a synthesis capability built on your own engagement history, your own client context, your own delivery patterns – one that converts AI tool usage into AI outcomes. The difference between using AI and capturing AI value is whether that infrastructure exists.

If you want to see what that looks like for a practice your size, email matthew@fieldway.org.

---

Related: Your Pipeline Is Healthy. Your Calendar Is Lying to You. | Why Your CRM's AI Doesn't Know When a Client Is at Risk