Your AI Works Fine. Your Company Is the Problem.

A new Publicis Sapient survey of 1,550 decision makers finds 73 percent of large enterprises use AI across most of their work, but only 10 percent say it is core to operations. Nearly half say the AI is already capable while their organization is not built to capture what it can do. The bottleneck has moved from the technology to the org chart, and most leaders are still trying to fix the wrong one.

There is a question I hear in nearly every executive conversation right now, and it always carries a note of frustration. We adopted AI, so why are we not seeing the value? The tools are in place. The pilots ran. The vendors got paid. And yet the business looks roughly the same as it did before. What went wrong?

This week, a major report answered that question with uncomfortable precision. Publicis Sapient released its 2026 Global Enterprise AI Report at VivaTech in Paris on June 17, based on a survey of 1,550 AI decision makers across six countries. The top-line finding: 73 percent say AI is used regularly or across most business processes, but only 10 percent say AI is core to how their business operates.

Sit with the size of that gap. Nearly three quarters of large companies have AI woven into daily work. One in ten has made it central to how the business runs. The distance between those two numbers is where billions of dollars of expected value is currently evaporating.

The technology is ready. The organization is not.

The most important finding in the report is not about adoption rates. It is about where the blockage sits. Forty seven percent of respondents said AI is already fully capable of meeting today's business needs. And 42 percent said the AI is capable, but their organizations are not set up to capture its value. More than one in five named the way their organization runs as the single biggest barrier to AI success.

Read that carefully, because it inverts the entire conversation most companies are having. Leaders keep asking whether the model is good enough. The people closest to the work are telling them the model is fine. The company is the problem.

This matches what I see on the ground. Most enterprises bolted AI onto a structure that was designed around human bottlenecks. The approval chains, the handoffs, the review cycles, the departmental boundaries, all of it was built to manage the limitations of human labor: people get tired, people need coordination, people can only do one thing at a time. When you drop a system that does not have those limitations into a structure built entirely around them, the structure becomes the new bottleneck. You made one part of the process infinitely fast and left it trapped inside a process that still moves at the old speed.

Nigel Vaz, the CEO of Publicis Sapient, put it directly: the enterprise was not designed for the speed, scale, and autonomy that AI makes possible. Deployment alone does not create advantage. The winners will be the companies that redesign how work gets done.

Why bolting AI on cannot work

There is a reason this approach fails so consistently. AI does not relieve a bottleneck unless you also remove the human-shaped constraints around it.

Imagine a content process where AI can now draft in seconds what used to take a writer a week. That sounds like a massive gain. But if the draft still has to wait three days for a manager's review, then route to legal for a five-day check, then sit in a queue for scheduling, you have not changed your output speed at all. You have an instant step surrounded by slow ones. The total cycle time barely moves. The AI is working perfectly. The system is wasting it.

This is why 42 percent of respondents can fairly say the AI is capable while their company cannot capture the value. Both things are true at once. The capability exists. The structure cannot hold it.

The report's regional breakdown makes the point sharper. In the U.S., the fastest adopters, 34 percent now say the way their organization runs is the primary constraint to AI success, not the technology. The further along a market gets in adoption, the more clearly it sees that the real work was never the tools. It was the redesign nobody wanted to do, because redesigning how work flows is harder, slower, and more political than buying a license.

The Hive Structure: the redesign that actually works

When I help companies move past this wall, I give them a simple model to organize around. I call it the Hive Structure, and it has exactly two roles. No more.

The first role is the bees. The bees are the AI, executing volume work at machine speed. Drafting, summarizing, generating, sorting, processing, anything that is high-volume and rules-bound. You want as much of this handled by bees as possible, and you want it handled fast, without a human in the middle of every step.

The second role is the beekeeper, sometimes the queen bee. This is the human, and the human's job is not to do the volume work. It is to direct the swarm, judge the output, and ensure the whole thing serves a purpose. The beekeeper decides what to make, sets the standard for good, and owns the decisions that carry real consequence.

The critical discipline is this: never add extra tiers. The instinct in most enterprises is to keep all the old management layers and just insert AI somewhere in the middle. That recreates the exact bottleneck you were trying to remove. If your redesigned process has bees, then a reviewer, then an approver, then a coordinator, then another approver, you have not built a hive. You have built your old org chart with a chatbot stapled to it, which is precisely what 42 percent of the survey is describing.

A real hive collapses the distance between generation and decision. The bees produce at volume. The beekeeper judges and directs. The work flows because there is nothing slow sitting between the fast parts.

The ambition gap is getting wider, not narrower

There is one more finding in the report that should worry every leader. The gap between ambition and readiness is not closing. In the U.S., 71 percent of respondents expect significant progress in scaling AI over the next 12 to 24 months, but only 20 percent say their organizations are fully equipped today to meet those expectations.

That is a 50-point gap between where companies expect to be and where they are actually built to go. And it exists in every market surveyed. Most organizations are moving faster in ambition than in execution, which is a polite way of saying they are writing checks their structure cannot cash.

This connects to a broader finding from other research this year. A separate study found that 79 percent of organizations face challenges adopting AI, a double-digit increase from 2025, and that 54 percent of C-suite executives admit adopting AI is tearing their company apart. The pain is not the technology. The pain is the organizational change the technology demands, and the fact that most companies tried to skip it.

The redesign is hard because it is political, not technical

If the fix is so clear, why do only 10 percent of companies get there? The honest answer is that the redesign threatens people, and the people it threatens are often the ones with the power to block it.

Every layer of approval and handoff in your current process exists because someone owns it. That manager who reviews the work, that coordinator who routes it, that approver whose sign-off is required, each of those steps is somebody's job, sometimes somebody's whole department. Collapsing the distance between generation and decision means asking some of those people to give up control they have spent careers accumulating. That is why redesign stalls. Not because anyone disputes the logic, but because the logic points at the org chart, and the org chart has feelings and budgets and political capital.

This is also why the redesign cannot be delegated to a project team and treated as an implementation detail. Restructuring how decisions flow is an act of leadership, and it requires someone with the authority to overrule the people whose roles are being compressed. When companies hand AI transformation to a working group with no power to change reporting lines, the working group produces a deck, the structure stays exactly as it was, and the company joins the 42 percent who say the AI is capable but the organization cannot capture it. The technology problem was solved by the vendor. The political problem was nobody's job, so nobody did it.

The leaders who break through treat the org redesign as the actual work, not the cleanup after the real work of buying tools. They start small precisely because small is where you can win a political fight: one workflow, one team, one beekeeper given genuine authority, and a measured result that becomes impossible to argue with. A proven win on one process is the only thing strong enough to move the people guarding the next one.

What to do, starting this week

Stop evaluating the model. The data is clear that the model is rarely your problem. Start evaluating the workflow around it.

Pick one process where you have already deployed AI and where you expected a big gain that has not materialized. Map the full cycle, every step from start to finish, and mark which steps are fast because of AI and which are slow because of human-shaped structure. You will almost always find the same picture: one or two instant steps, surrounded by a queue of approvals and handoffs that were designed for a slower world.

Then redesign that one process around the Hive Structure. Let the bees own the volume. Give a single beekeeper the authority to judge and direct, and remove the tiers that exist only to manage human limitations the AI does not have. Do not try to transform the whole company at once. Prove it on one workflow, measure the difference, and use that as the template. One redesigned process that demonstrably moves faster and costs less becomes the reference point for the next, and the next, until the redesign stops feeling like a threat and starts feeling like the obvious way to work.

The companies in that 10 percent who say AI is core to operations did not buy a better model than everyone else. They did the redesign everyone else postponed. So here is the question to bring to your next leadership meeting: are you still trying to fix your AI, when the thing that actually needs fixing is the shape of your company around it?

Sharon Gai is an AI transformation strategist, keynote speaker, and author of How to Do More with Less Using AI. She advises Fortune 500 companies on AI adoption and organizational redesign.