23 June 2026

AI has moved from experiment to infrastructure

For years, AI in construction lived in the innovation department: pilots, demos, conference slides. That phase is over. The reliable signal is no longer what the technology can do, it is what serious businesses are doing about it operationally.

Look at where the money and the org charts are moving. Suffolk, one of the largest contractors in the US, has publicly committed tens of millions of dollars to AI and now embeds dedicated AI engineers on its jobsites. Skanska has built an internal suite of AI assistants, its Sidekick tools, used across the group, including a safety-focused assistant rolled out across its US operations. In the UK, Skanska has stood up a digital academy with structured data apprenticeships and AI training cohorts for its staff.

On the tools side, reality-capture platforms such as OpenSpace now operate across more than a hundred thousand projects worldwide, quietly turning site photography into structured progress data. That is not a pilot. That is infrastructure.

Why this matters more than the demos

None of those decisions were made for a press release. Contractors do not fund AI teams, apprenticeship schemes and group-wide assistants unless the work is paying for itself. Operating decisions are the honest indicator, because they are expensive to reverse.

And they share a theme worth noticing: the leaders are not primarily buying robots or moonshots. They are investing in people, data and the unglamorous connective tissue, training, records, assistants over their own information. The competitive gap this creates is structural. A business that burns less skilled time per pound of output does not just have better margins; it can bid more, respond faster and keep its best people doing work they actually value. That compounds, and it cannot be closed later by working harder.

A note on numbers

You will see spectacular statistics in this space: named contractors, precise percentages, millions saved. Some are real. Many do not survive contact with a primary source. Before this article was published, we fact-checked a set of widely circulated claims and cut the majority because no first-hand evidence exists for them.

We would rather give you three verifiable facts than ten impressive ones. It is the same standard we apply to our own work: if a saving cannot be measured in your accounts, we do not claim it in our reports. Treat any AI pitch that leads with unattributed percentages accordingly.

The SME advantage

Here is the part that gets missed. The capability driving the Tier 1 programmes, frontier models, automation tooling, private assistants over your own documents, is available at SME prices. What the giants buy with headcount, a smaller firm can buy with focus.

Smaller firms also hold a genuine structural advantage: agility. You can decide, deploy and adapt in weeks, without a steering group in sight. The constraint is not access to the technology. It is knowing exactly where it moves your bottom line, and not wasting capital on the wrong tools.

In our experience the answer is rarely glamorous. It is bid production, reporting, site records and compliance admin, the work that quietly eats your best people's weeks. Start there, measure honestly, and expand what proves itself.

The firms treating AI as infrastructure are pulling away from the firms still treating it as a curiosity. You do not need a Tier 1 budget to be in the first group. You need a first workflow, a measured result, and the discipline to build from there.