AI didn’t change how your company works. That’s why it isn’t working.

McKinsey set out to answer a specific question: out of everything a company could change when it adopts generative AI, what actually moves the number that matters — EBIT? They tested 25 organizational attributes against real outcomes, for organizations of every size. One factor came out ahead of all the others: the redesign of workflows has the biggest effect on an organization's ability to see EBIT impact from gen AI. Not model choice. Not which vendor got the contract. The way the work itself gets restructured around the tool.

Here's the problem. Only 21% of organizations using generative AI have fundamentally redesigned even one workflow around it. The other roughly four in five have done the easier thing: dropped a tool into the existing process and asked people to use it when convenient.

The gap between "using AI" and "AI working"

That easier path explains a pattern a lot of executives have noticed and not known what to call: the company is using AI. Licenses are active. Adoption numbers look fine on a dashboard. And still nothing shows up in the P&L.

Gallup's Q1 2026 survey of 23,717 employed U.S. adults found the same gap from the inside. In organizations that have adopted AI, 65% of employees say it's improved their productivity and efficiency — regardless of how often they personally use it. That's a real, widely felt signal. But ask a sharper question, whether AI has transformed how work actually gets done, and the number collapses: only about one in ten employees strongly agrees.

Put those two numbers next to each other and you have the diagnosis. People feel faster. The organization isn't different. A tool bolted onto an unchanged process can genuinely save someone twenty minutes on a task and still leave the process — the handoffs, the approvals, the org chart — exactly as it was. Twenty minutes saved by one person, absorbed into slack time, doesn't show up as a line on anyone's income statement.

What "layering on" looks like in practice

This is the same failure mode we described in why most AI pilots never reach the P&L: flawed workflow integration. Picture what that looks like on the ground.

• A drafting tool speeds up writing the proposal, but the proposal still routes through the same five approvals it always did.
• An assistant summarizes call notes, but the sales process built around those notes hasn't moved.
• A chatbot answers customer questions faster, but nobody redesigned who owns the escalations, so the bottleneck just relocated.
• A team gets a slicker intake form, but everything downstream still gets processed the old way, one at a time.

Each of these is a real improvement at the task level. None of them is a redesign at the workflow level. That distinction is the entire finding.

What redesign actually means

Redesign isn't a bigger rollout. It's a smaller, harder decision: pick one workflow and change its shape — who touches it, in what order, with which handoffs removed — instead of inserting AI into the shape that already exists. That usually means fewer steps, not more tools. Fewer handoffs, not a longer software list. A process that looks different on a whiteboard, not one that just runs faster inside the same boxes.

It's a harder sell internally than buying a license, because someone has to own changing how a team works, not just approve a new subscription. That's precisely why only one in five companies have done it. It's also why the effect shows up so clearly in McKinsey's data and nowhere else.

The honest starting point

You don't find the workflow worth redesigning by guessing, and you don't prove a redesign paid off by feel. That's the order we work in: map where AI would actually change the shape of the work, not just speed up a task, in an AI Opportunity Audit — then rebuild one real workflow and measure it against a baseline in a Pilot Sprint. The redesign is the hard part. It's also the only part the data says is worth doing.

---

Source: McKinsey, "The State of AI: How organizations are rewiring to capture value" (March 2025), and Gallup, "Rising AI Adoption Spurs Workforce Changes" (Q1 2026). If your team feels the productivity gain but can't point to it on the P&L, why most AI pilots never reach the P&L covers the same gap from the pilot side — or take the AI Ownership Scorecard to see where your own workflows stand.