The CM Awards Shortlist Shows What Real AI ROI Looks Like in UK Construction

The Construction Management Best Use of AI shortlist for 2026 is the most practically useful document in UK construction AI right now. Not because it tells you what AI can do in theory - because it tells you what AI is doing, right now, in named UK firms, with documented results.

Six entries made the shortlist. Every one has measurable outcomes:

BidLevel / ProcurePro - AI extracts and standardises vendor quotes from tender returns, reducing quote-levelling time by 80%. Roughly 60 seconds per package, down from what would previously take hours of manual comparison work. The commercial team gets consistent, structured data from unstructured tender documents without touching a spreadsheet.

Hoppa AI / AtkinsRealis - Natural language processing applies ISO 19650-compliant document classification at scale. On a nuclear decommissioning project, classification time dropped by 90% and ran 100 times faster than manual workflows. The volume of documentation in nuclear and infrastructure projects makes this kind of capability not just useful but essential.

Knowledge Hub AI / Willmott Dixon - Claude-powered (Anthropic) chat interface built specifically for bid support and content generation. Saves 28-plus hours per bid. A nine-question bid response drafted in four minutes and 37 seconds. The Willmott Dixon entry is notable for explicitly naming Anthropic Claude as the underlying model - a signal that frontier AI is now operating directly inside major UK contractor operations.

NavLive - LiDAR handheld scanner generating RICS-grade surveys, 2D floorplans, and 3D point clouds directly on site without a cloud upload. Saved approximately £50,000 on a single project versus traditional survey methods. Already adopted by AWW Architects specifically for Building Safety Act compliance. Projects that previously required days of surveying are completing in under an hour.

Auto ITP / Taylor Woodrow - NLP generates inspection and test plans from project specifications automatically. 50% reduction in drafting time. Estimated annual cost avoidance of £474,700.

District One / Ramboll - Generative design for district energy networks. Network routes that previously took weeks to produce are taking hours. £32 billion in potential national savings modelled if the approach is deployed at scale.

None of these are proof of concepts. None are in a lab. They're on live projects at named UK firms, producing documented commercial and operational outcomes.

The Building Safety Act's Golden Thread requirement - a live, queryable digital record of all building decisions across a structure's lifecycle - is now being actively enforced by the Building Safety Regulator. The shift is significant: information that can't be navigated or queried by an accountable person is increasingly being treated as incomplete, regardless of whether it was technically submitted.

AI is playing a dual role here. On the positive side, AI tools are automating document classification, compliance tagging, and record organisation - making it easier for firms to get their Golden Thread into shape. Hoppa AI's work with AtkinsRealis at scale (covered above) is a direct example of AI applied to exactly this kind of document management challenge.

But AI is also exposing problems. Systems auditing existing records are identifying inconsistencies, gaps, and classification errors that organisations assumed were fine. This is the uncomfortable flip side of AI-assisted compliance. The tools that help you fix your records are the same tools that can find what's broken in them.

The practical implication for UK contractors and consultants is direct. The BSR is not asking whether digital records were submitted. It's asking whether they can be used - whether a building safety manager or accountable person can actually navigate and query the information when they need it. Static document dumps that satisfy a tick-box but can't be interrogated are no longer adequate.

The ServiceTitan 2026 Commercial Specialty Contractor Industry Report - drawn from a survey of over 1,000 construction leaders - finds that 38% of contractors now report measurable business impact from AI. That's up from 17% in 2025. More than doubled in a year.

The top use cases are cost estimation and budgeting (24% of firms) and bid management (22%). These are core commercial functions, not experimental applications. Firms are using AI where the financial exposure is highest and where getting it wrong costs real money.

The market context matters. Seventy-one per cent of contractors report rising labour costs, up from 55% last year. Only 20% operate on a single integrated technology platform. Strong backlogs (41% of firms hold 12-plus months of secured work) mean demand isn't the problem. Margins are. And AI is the lever being pulled to protect them.

The 62% of contractors who aren't yet reporting measurable AI impact deserve a closer look too. Some will be actively deploying and not yet measuring. But many are watching. And in a market where the early adopters are compounding their advantage with every passing quarter, watching is a losing strategy.

Autodesk has renamed its Construction Cloud platform as Autodesk Forma and shipped over 70 product updates in its March 2026 release. The most significant is the AI Project Data agent exiting beta. Teams can now surface project insights, automate routine tasks, and query project data directly using AI embedded within the platform.

The product line rename brings everything under the Forma umbrella: Autodesk Build becomes Forma Build, Autodesk Docs becomes Forma Data Management, BIM Collaborate Pro becomes Forma Design Collaboration, Autodesk Takeoff becomes Forma Takeoff. Three simplified hub-level roles ease onboarding.

For the large number of UK construction firms using Autodesk products, the AI Project Data agent is available now. It's worth understanding what it can actually do in practice - because the gap between what's been shipped and what's being used on most projects is still significant.

The most significant trend across UK construction AI coverage this week is the shift from agentic AI as a concept to agentic AI as a live deployment. The emerging pattern: AI agents that can read drawings, check submittals, analyse risk clauses in contracts, and surface schedule conflicts are now operational at several major UK contractors.

The architecture that's taking hold is multi-agent orchestration - systems where one agent plans, another executes, another reviews. For construction, the practical translation is AI that can own a complete workflow thread: flagging a delay, pulling the relevant contract clause, drafting the variation notice, logging it to the CDE, all without a human stitching each step together.

This is the direction Buildots illustrates at a more accessible level. Its hardhat-mounted 360-degree cameras compare daily site imagery against BIM models and project schedules, automatically detecting discrepancies. Claimed reduction in project delays: up to 50%. The platform now integrates with both Procore and Autodesk Construction Cloud, making it a plug-in layer on top of existing tools rather than a replacement.

UK Construction Week London (12-14 May) is running a dedicated Contech and AI Hub stage for the first time. Over 150 hours of CPD content on AI, with confirmed speakers including David Philp from Bentley Systems and Professor Sean Smith from the University of Edinburgh on building AI-ready teams.

The event draws more than 25,000 construction professionals. The fact that AI gets its own dedicated stage - not just a session within a broader programme - is itself a signal. It reflects where industry investment attention and professional development demand are landing.

If you're planning to attend, the AI Hub content is worth prioritising. The combination of practitioners sharing live deployments and researchers setting direction should give a clearer picture of where the industry sits in 2026 than most internal briefings will.

Sumitomo Heavy Industries and NEC Corporation have announced a joint initiative to develop an AI near-miss detection system for construction sites. The system analyses camera footage and sensor data from excavator-mounted cameras, automatically flags risk scenarios, and generates near-miss reports with concise summaries. Practical deployment is targeted for fiscal year 2027. A proof of concept in September 2025 successfully identified accident scenarios from live footage.

The significance is the approach: embedding detection into the machine itself rather than relying on fixed CCTV coverage. Excavators generate blind spots and coverage limitations that fixed cameras can't resolve. An AI system reading data from the excavator's own sensors addresses that directly.

Near-miss reporting is also chronically underused on UK sites - the effort involved in documenting incidents means many go unrecorded. AI-generated summaries requiring no manual input from site staff could change that meaningfully.