Kier's 10,000-user Copilot rollout offers a lesson for construction

Kier’s Microsoft 365 Copilot rollout may look quieter than the week’s robotics clips or frontier model benchmarks. Its importance comes from something more durable: a tier-one contractor pushing AI into normal work at enterprise scale.

The UK contractor has put Copilot into its everyday working environment as part of a three-year agreement with Microsoft, with the wider package including Teams Premium licences for its 10,000-strong workforce. The important detail is the distribution pattern. Kier is pushing AI beyond a small digital team and into the normal operating layer of the business, including knowledge workers and field-based roles.

That is the shift our sector keeps talking around. For years, construction has been good at pilots and poor at repeatability. The gap between a useful trial and a scaled operating model is where most digital initiatives go to quietly expire. Kier’s announcement is interesting because it points at the less glamorous work of adoption: training, confidence, governance, licence coverage and workflow redesign.

Stuart Togwell, Kier’s chief executive, framed the move in operational rather than novelty terms:

That quote is useful because it avoids the usual AI theatre. The language is about work, service and outcomes. The phrase 'Naturally Digital' also matters. It implies that digital should not sit as an accessory to delivery. It should be embedded in the way people brief, decide, coordinate and report.

The adoption signal is stronger than the product signal. Copilot is now a known quantity. The story is that a tier-one UK contractor has decided the productivity upside is mature enough to place inside daily work at scale. A rollout to 10,000 employees, backed by a three-year Microsoft agreement, suggests AI is being treated less as a pilot and more as operating infrastructure.

The details reinforce that view. Copilot is being embedded into everyday workflows rather than confined to a handful of specialist use cases. Teams Premium coverage extends AI-enabled collaboration across the workforce, reaching the meetings, information exchanges and coordination tasks where delivery friction often accumulates. Kier also reports that AI confidence rose to 87 per cent following training programmes, indicating that adoption is advancing alongside capability rather than being left to chance. Taken together, the rollout, training effort and long-term commitment point to an organisation preparing for sustained AI use rather than a short-lived experiment.

The construction angle is obvious. Our sector loses huge amounts of time in document handling, coordination, meeting follow-up, information retrieval, reporting and repetitive admin. Autodesk and FMI previously found that construction professionals spent 35 per cent of their time, more than 14 hours a week, on non-productive activities such as looking for information, resolving conflicts and dealing with mistakes and rework. Those are the work Copilot class tools best placed to attack first.

This does not mean Copilot is suddenly designing bridges or resolving contractual disputes. It means the first useful wave of enterprise AI in construction may be much more ordinary: summarising project correspondence, drafting meeting actions, finding buried information, preparing first-pass reports, cleaning up internal communications and helping people move faster through the administrative sludge around delivery.

Office productivity stories are easier. The tougher question is whether AI improves work at the edge of delivery, where information is messy, time is short and decisions are constrained by the physical reality of a project. Kier’s reference to field-based roles, therefore, deserves attention. If Copilot becomes useful to site teams, supervisors and operational managers, the value case changes. AI stops being a knowledge worker convenience and starts becoming a delivery support layer.

Theo Michalopoulos, General Manager for Public Sector at Microsoft UK, made the strategic framing explicit:

The phrase 'secure, AI-powered tools at scale' is doing a lot of work. For major contractors, security and scale are inseparable. A useful AI tool that cannot survive information governance, client confidentiality, project control discipline and operational risk management is not really enterprise-ready. The market is moving beyond whether an individual can get value from a chatbot. The question is whether an organisation can deploy AI without creating a shadow IT problem, a data leakage problem or a quality assurance problem.

The RICS AI in Construction report found that AI was the leading construction technology for increased investment in 2025, with 56 per cent of surveyed investors planning to allocate more funds to AI than the previous year. Yet the same research showed that implementation remains immature. Around 45 per cent of respondents reported no AI implementation, another 34 per cent were still in early pilots, and less than 1 per cent reported fully embedded organisation-wide AI use.

That is the gap Kier is stepping into. The clear indication is that the competitive frontier is moving from interest to institutional capability. Firms that can train people, redesign processes, govern usage and measure value will move faster than firms that simply buy licences and hope for behaviour change.

There is also a strategic timing point. Kier made three senior digital and technology appointments earlier this year, and its announcement references digital construction work on projects including HMP Five Wells, HMP Millsike and the Ministry of Justice Digital Twinning Project. In other words, Copilot is not being presented as a standalone productivity add-on. It sits inside a broader digital construction narrative that includes data, automation, drones, robotics, apprenticeships and immersive safety training.

That is the right pattern. AI will not transform project delivery as a disconnected layer. It becomes valuable when linked to the systems that already carry project information, commercial risk, programme logic, safety observations and client commitments.

• Treat AI rollout as a delivery transformation programme, not a software deployment. The licence is the easy part. The harder work is training, workflow design, confidence building and governance.

• Start with the work people actually do. Meeting actions, correspondence, reporting, document search and knowledge retrieval may sound mundane, but that is where large time savings accumulate.

• Measure adoption quality as well as usage. Kier’s AI confidence figure is a useful reminder that productivity depends on whether people feel able to use tools on real tasks.

• Bring site teams into the design early. If AI only helps head office, the sector will miss the larger prize of improving coordination between office decisions and physical delivery.

• Link AI to existing digital construction capability. The firms that win will connect Copilot class tools to data environments, project controls, safety systems and client reporting, rather than treating them as a separate productivity toy.

For Project Flux readers, Kier’s announcement is a useful benchmark. This is what scaled AI adoption in construction is beginning to look like: less demo, more operating model. The next question for every contractor and consultant is simple. Where is AI already inside everyday work, and where is it still stuck in the pilot folder?

Subscribe to our newsletter for weekly insights on how AI is affecting businesses.

Kier press release

Construction Wave coverage

RICS AI in Construction report

Balfour Beatty Microsoft AI investment

Construction Disconnected report PDF

All content reflects our personal views and is not intended as professional advice or to represent any organisation.