Why Only 6% of Companies Actually Win at AI

Quick answer

Why Only 6% of Companies Actually Win at AI explains what the change means for UK SMEs and how to turn it into a practical next step. The process is to identify the business decision, connect the data, then automate only the parts that improve speed or reliability.

I see this all the time with UK SMEs. They've had a play with ChatGPT, maybe built a Power Automate flow or two, stuck "AI-driven" on the website. Feels like progress. Then someone asks "so what's it actually saved us?" and the room goes quiet.

McKinsey's latest State of AI report lays it bare. 88% of organisations are now using AI regularly, but only 6% qualify as genuine high performers getting real bottom-line impact. The rest are stuck in what they call "pilot purgatory": lots of experiments, nothing scaled, nothing that moved the needle.

So what are the 6% doing differently?

Here's the bit that caught my eye. The gap between winners and everyone else is not about who has the fanciest models or the biggest budgets. It is about how they approach the work:

• They redesign processes, not just digitise them. 55% of the high performers fundamentally rework how things get done when they deploy AI. The rest just slap AI on top of broken processes and hope for the best.
• They pick two or three things and go all in. No scattergun pilot approach. Concentrate fire on the highest-impact use cases and commit to full-scale deployment.
• Leadership actually owns it. Nearly half of the high performers report strong executive ownership and long-term commitment. Only 16% of the rest can say the same.

McKinsey put it plainly: "The dividing line between AI leaders and laggards is no longer technical access. The true differentiator is organisational plasticity." Basically, can your business actually bend and adapt, or is it rigid?

Why this matters for SMEs especially

Big corporates can afford to waste money on AI pilots that go nowhere. SMEs can't. Every pound matters. And the data from S&P Global is brutal: 42% of companies abandoned most of their AI initiatives in 2025, up from 17% the year before. That is a proper nightmare if you've sunk budget into something that never delivered.

The pattern I see over and over is the same. Someone gets excited about AI, buys a tool, builds a proof of concept, shows it to the board, everyone nods, and then it dies. Not because the tech failed, but because nobody redesigned the actual process the AI was supposed to improve. It's like buying a Ferrari and driving it round a car park.

Three things to actually do

1. Stop experimenting and pick a lane. Find the single process that eats the most time or causes the most errors. Customer reporting, invoice processing, stock reconciliation, whatever it is. Pick one. Commit to automating it properly, end to end.

2. Redesign before you automate. Map the process as it actually happens now, not how the manual says it should. Cut the waste first. Then layer the AI on top of something lean. If you automate a mess, you just get a faster mess.

3. Measure something real. Hours saved, error rate dropped, turnaround time halved. Not "we tried AI." Actual numbers. If you can't measure it, you can't justify the next investment.

The 6% are not smarter or luckier. They just do the boring groundwork that everyone else skips. Sort the process, commit to scale, measure the result. Not glamorous, but it works.