The 11 Percent: Why Most Businesses Are Deploying AI Backwards

KPMG just finished surveying global organisations about their AI programs. The headline figure: companies are planning to spend a weighted average of $186 million on AI over the next 12 months. The buried figure: only 11 percent of those organisations have deployed AI in ways that produce enterprise-wide business outcomes.

That gap — between what's being spent and what's being achieved — is not a technology problem. It's a sequence problem. And if you're a business owner trying to figure out why your AI tools feel underwhelming, this data tells you exactly why.

The Numbers That Should Worry Every Business Owner

The KPMG Global AI Pulse survey breaks companies into two groups: "AI leaders" (organisations actively scaling agentic AI with measurable results) and everyone else. The performance gap between them is striking.

Among AI leaders, 82 percent report that AI is delivering meaningful business value. Among their peers, that number drops to 62 percent. A 20-point spread might sound modest until you understand what's underneath it.

It's not that one group spent more. The laggards are spending enormous sums. It's that the two groups deployed AI in fundamentally different order. AI leaders redesigned their processes first, then deployed AI to operate within those redesigned structures. Everyone else did the reverse: they bought tools, layered them onto existing workflows, and wondered why the returns were underwhelming.

In IT and engineering, 75 percent of AI leaders use agents to accelerate code development, versus 64 percent of peers. In operations and supply chain, it's 64 percent versus 55 percent. These aren't marginal tool adoption differences — they reflect entirely different philosophies about what AI is actually for.

The Layering Mistake Almost Every Business Makes

Here is what the losing approach looks like in practice: a company buys a ChatGPT license. Then adds a summarization tool for meeting notes. Then purchases a CRM copilot. Each tool solves a local problem. Each tool produces incremental gains. After 18 months, the company has spent significantly and has a collection of useful-but-disconnected tools that haven't moved the needle on headcount, revenue, or margins in any material way.

KPMG calls this "layering models onto existing workflows." A copilot here. A summarization tool there. The problem is that the underlying workflow — the sequence of steps, decisions, handoffs, and bottlenecks — is the same broken process it always was. You've just made individual steps within a broken process slightly faster.

The 11 percent doing this right ask a different question before they buy anything: "If we were starting this process from scratch, knowing what AI can do, how would we design it?" The answer to that question is almost never "exactly like we do it now, but with a summarization tool at the end."

What "Process First" Actually Looks Like

Concrete example. A mid-size professional services firm has a proposal generation process that takes their senior consultants an average of 14 hours per engagement. The typical fix: buy a proposal-writing AI that cuts drafting time by 40 percent. Save six hours. Call it a win.

The process-first approach looks different. Start by mapping the entire proposal workflow: intake call, discovery, competitive research, pricing, drafting, review cycles, and client presentation. Then ask where decisions are being made manually that don't require judgment — where a human is essentially pattern-matching to previous proposals. Then build AI into those specific decision points, not just the writing step.

Done right, the same firm gets proposal generation down to under three hours, eliminates one full review cycle, and captures the pricing logic in a system that gets smarter with each engagement. The headcount implications are real: two senior consultants can now carry the proposal load that previously required four.

That's not incremental. That's a different business.

The KPMG data bears this out in a counterintuitive place: governance. Among organisations still in the experimentation phase, only 20 percent feel confident in their ability to manage AI-related risks. Among AI leaders, that number rises to 49 percent. But here's the insight: the leaders didn't get confident by slowing down and being cautious. They embedded governance into the deployment pipeline — monitoring outputs, building escalation paths, maintaining audit trails — and that infrastructure is what allows them to move faster, not slower. Governance isn't a brake. For the companies winning right now, it's the accelerator.

Why Smaller Businesses Have an Unexpected Advantage

Here's what the KPMG data doesn't say explicitly, but implies: the companies worst positioned to execute the "process first" approach are large enterprises with deeply entrenched workflows, legacy ERP systems, and compliance layers that make process redesign a multi-year initiative.

If you're running a company with 20 to 200 employees, you don't have that problem. Your processes are malleable. You can redesign a proposal workflow this quarter and have it operational next month. You can rebuild a client onboarding process without navigating three layers of departmental approval and a six-month IT procurement cycle.

The 11 percent at the enterprise level are the ones who figured out how to overcome their own inertia. At the SMB level, inertia is much easier to overcome. You have a structural advantage that most business owners don't recognise because they're comparing themselves to what big companies have, rather than what big companies can actually execute.

The window for that advantage is not unlimited. As more SMBs go through proper AI implementation, the gap between those who've done it right and those who haven't will compound in revenue, margins, and the ability to take on more clients without proportional headcount growth.

Where to Start

The right starting point is not "which AI tool should we buy." It's "which of our processes, if redesigned with AI in mind, would change our unit economics most?" That question requires knowing what AI can actually do in your specific context — not in theory, but in practice, with your data, your workflows, and your team.

That's the work we do at JR Intelligence. We start with an AI audit that maps your current operations, identifies the processes with the highest redesign leverage, and gives you a specific implementation roadmap — not a list of tools to buy. Because the companies that end up in the 11 percent don't start by buying. They start by thinking.

If you're ready to do it in the right order, visit our services page or get in touch.