Week 14: The 36% vs the 31% - and the Infrastructure Gap in Between

The biggest number this week came from the Association for Project Management. Seventy-five per cent of construction project professionals now use AI. Two years ago it was 15%. That's not incremental. That's a category shift.

But here's the thing about adoption statistics. They measure who's using something. They don't measure how well, or to what end, or whether the results justify the investment. The same APM survey found that 36% of firms have a clear strategy delivering tangible benefits. Another 31% are still gathering proposals for pilot projects. Those two groups are both "using AI." They're having completely different experiences of it.

The CM Awards shortlist this week put numbers on what the first group looks like. Six firms, six real deployments, all with documented outcomes. Ninety per cent time savings on document classification at AtkinsRealis. Twenty-eight-plus hours saved per bid at Willmott Dixon. £474,700 in estimated annual cost avoidance at Taylor Woodrow. NavLive saving roughly £50,000 on a single survey project.

These aren't productivity experiments. They're commercial results. And the gap between the firms producing them and the 31% still in the pilot-gathering stage is growing every week.

Thursday's brief added an important qualifier. Bluebeam's research found only 11% of AEC firms describe themselves as fully digital across all project phases, even though 84% plan to increase technology investment this year. Archdesk's agentic AI analysis arrived at the same uncomfortable conclusion from a different angle: AI is only as good as the data feeding it. Firms with fragmented systems - estimating in one tool, scheduling in another, costs in a spreadsheet - will see limited benefit from AI investment regardless of which tools they buy.

The AI adoption story in UK construction right now is actually two stories. The first is the 36%: firms that have cleaned up their workflows, structured their data, and deployed AI in core commercial and compliance functions. They're compounding their advantage every quarter. The second is everyone else: using AI in some form, but not yet at the level where it changes outcomes.

The week also threw up a system-level problem. The UK government has a 6GW AI data centre target for 2030. The current trajectory gets to 4.9GW - and only if a wave of planning consents clears by 2027. Planning, not funding or technology, is the constraint. That's a construction and planning tech problem as much as it is a policy one. The firms that understand it are well placed.

There was also good news for smaller firms. Construction AI launched an AI-native project management platform built by a Chartered Builder with no software background - 700,000 lines of code, 60-plus AI tools, priced at £2,000 lifetime. And Re-flow's new PAS 2080 module embeds carbon compliance into the workflows infrastructure contractors already use. The tools are reaching firms of every size now. The barrier to entry has dropped.

The message from the week is consistent. Adoption has surged. Results belong to those with a strategy. The workflow and data foundation determines whether the tools work. And the infrastructure the UK needs to remain competitive depends on construction firms moving faster than the planning system currently allows.

The question isn't whether to invest in AI. It's which side of the strategy gap your firm is on.