From experiment to expectation
Two years ago, the conversations were about proof of concepts. Small teams with dedicated budgets, sandboxed from the real work of the business. The goal was to prove that AI could work. Not to change how anything actually ran.
That era is ending. According to McKinsey's 2025 State of AI report, 78% of organizations now use AI in at least one business function, up from 55% just a year earlier. The question has shifted from "should we try this?" to "why isn't this working yet?"
And that second question is more important than it sounds. Because a lot of companies that started with pilots never made it through to the other side. They ran the experiment, saw promising results, wrote the report. Then the initiative quietly stalled. The tool sat unused. The team went back to their regular work.
The pilot trap
This isn't a failure of ambition. It's a failure of integration.
AI that lives beside your workflow doesn't change your workflow. It just adds another thing to check. A dashboard nobody opens. A tool that requires a separate login. A process that runs in parallel with the old process until someone decides which one to trust.
The companies that are actually moving forward share one thing in common: they stopped treating AI as a separate initiative and started embedding it into how work actually gets done. The AI isn't a product you evaluate on a quarterly basis. It's how you run your approvals, how you process your documents, how you schedule your operations.
PYMNTS reported in 2025 that companies are consolidating "disconnected tools into unified AI platforms designed to support core workflows, control inference spending, and operate reliably across the organisation." The fragmented pilots of 2023 are being replaced by platforms that sit inside the systems people already use every day.
Still running the same experiment?
If your AI project has been in pilot mode for more than three months, it's time to talk.
What the numbers actually tell us
62% of enterprises are now actively building or experimenting with AI agents, with more than half having agents already in production. 39% are running ten or more agents simultaneously. These aren't experiments. They're operations.
But here's the number that matters most: only 6% of organisations report success across all key dimensions: AI fully embedded enterprise-wide, no significant barriers to scaling, and technology investments delivering expected results. There is an enormous gap between using AI and actually transforming how you work.
That gap is where companies are winning and losing right now.
What this means for established companies
If you run an established business with real operations, real staff, and real processes built over years, this moment matters more than you might think.
The companies catching up fastest right now aren't the startups. They're the mid-sized manufacturers, distributors, and professional services firms who decided to go all-in on embedding AI into the actual flow of work. Not as a side project. Not as an IT initiative. As the new way the business runs.
The mistake to avoid is launching another pilot. The opportunity is to take something that already works at small scale and ask a harder question: what would it take to make this how we actually operate?
The pilot phase is over. The companies that figure that out first are the ones that will be hardest to catch.