Earlier this year, OpenAI's chief operating officer said something that cut through months of AI hype: "We have not yet really seen AI penetrate enterprise business processes." This was not a critic or a sceptic. This was the person running commercial operations at the company with arguably the most to gain from enterprise AI adoption. The candour was notable. So was the gap it described.
The picture that has emerged across the first quarter of 2026 is of an enterprise AI market that is spending heavily, piloting broadly, and transforming much less than either the vendors or the buyers had hoped. AI budget commitments from large companies are real and significant. The gap is between budget and change: money is flowing, but the business processes at the end of the pipeline remain largely untouched.
What is blocking the translation? A few patterns keep appearing. First, integration complexity. Enterprise systems are not clean APIs and structured databases. They are a tangle of legacy software, paper-based workflows, regulatory constraints, and human judgment calls that resist the kind of clean automation that works well in demos. An AI that handles structured queries on a tidy dataset performs very differently in a real accounts payable department with 30 years of exception-handling quirks baked into informal processes.
Second, the locus of change. AI tools are landing in the hands of individual workers, not in the hands of the people redesigning processes. A lawyer using a legal AI assistant has genuinely changed how she works. But the firm's billing model, intake workflow, and quality review process remain as they were. Individual productivity gains and organisational transformation are different things, and most enterprise AI deployment is producing the former, not the latter.
OpenAI's response to this gap has been to bring in the consultants, announcing in February that it was partnering with management consulting firms to help enterprises actually deploy AI at scale. The move is a reasonable acknowledgement that selling software is not enough when the blocker is organisational change management. But it also signals that the AI industry is entering a more mature, and more difficult, phase than the initial tool-selling model suggested.
The VC surveys from late 2025 captured this tension well. Enterprise AI budgets for 2026 were projected to rise, but through fewer vendors: consolidation driven by companies demanding proof of ROI rather than running multiple pilots indefinitely. One MIT survey found that 95 percent of enterprises were not yet seeing meaningful returns. That number has likely improved through early 2026, but the direction of travel in analyst commentary has shifted from "when will enterprises adopt AI" to "why is the adoption so much harder than expected."
None of this means enterprise AI is failing. It means it is in its awkward phase: past the excitement of early pilots, not yet at the point where transformation is visible in productivity statistics or org charts. The companies that have done well are generally those with clear, narrow use cases, strong technical teams, and willingness to redesign workflows rather than just layer AI tools on top of existing ones. That combination is less common than the vendor pitch decks suggest. Getting from pilot to process change is where the real work is, and most organisations are still figuring out how to do it.