Construction AI: UKCW opens, procurement gets paid, and agents earn their security questions

UK Construction Week London begins this morning at ExCeL and runs through Thursday 14 May. Expected attendance is 25,000+ across 14 stages, with over 150 hours of CPD-accredited seminars. The standout addition for 2026 is the ConTech & AI Hub - sessions span automation, robotics, digital twins, AI-driven decision-making, data-intelligent systems, generative design, on-site risk anticipation, automated estimating, and AI in construction sales and marketing.

If you can only do one day, today (Tuesday) and tomorrow (Wednesday) have the densest AI programming. Send your operational leads, not just the digital team - the value is in the procurement-AI and project-controls conversations, which need delivery-side decision-makers in the room.

ProcurePro - the Brisbane-founded but globally pitched procurement-AI platform - announced an US$11m (A$15m) raise this week, led by Queensland government-backed QIC Ventures, with existing investors AirTree and Glitch Capital topping up and French main contractor Bouygues joining the cap table via Paris-based VC ISAI. The round values the six-year-old company above US$100m. ProcurePro has run more than 6,000 construction projects globally and will use the capital to hire 100 staff across product, engineering and go-to-market, with UK, Middle East and US expansion explicit.

The platform's pitch - consolidating 20+ fragmented procurement workflows so main contractors can compare, select, approve and contract on one AI-powered system - is procurement's version of what BIM tried to do for design. The Bouygues move from customer to investor is the credibility signal here.

UK quantity surveyor Sofoclis Patsalos has been visibly pushing MyQS.ai on LinkedIn this week as a practical answer to AI in QS work. The platform takes drawings, voice notes or text and produces a full NRM1-tagged Bill of Quantities - labour and materials split, no duplicates - in minutes rather than days. The standout design choice is transparency: MyQS explicitly flags what it has excluded (perimeter fencing, vegetation clearance, kitchen units and similar) rather than hiding hallucinations behind a clean output.

That transparency choice matters. PI insurers and clients are increasingly asking what the AI did and did not consider; tools that volunteer this information are more defensible than tools that present a polished number with no audit trail.

OpenAI shipped GPT-5.5 Instant on 5 May as the new default model in ChatGPT, replacing GPT-5.3 Instant. The headline numbers are the ones to keep on hand when briefing colleagues: 81.2 on AIME 2025 (up from 65.4), 76 on MMMU-Pro (up from 69.2), and - most relevant for construction - a 52.5 per cent reduction in hallucinated claims on high-stakes prompts covering law, medicine and finance.

That hallucination delta is the line worth quoting when defending AI use in contractual, technical or compliance contexts. It is also a useful reminder that the model-quality bar is still moving every quarter, so any vendor benchmark older than two months should be re-tested.

Palisade Research published "Language Models Can Autonomously Hack and Self-Replicate" on 7 May. The paper tested OpenAI's GPT-5, 5.1, 5.4; Anthropic's Claude Opus 4, 4.5 and 4.6; and Alibaba's Qwen against systems with deliberately planted security flaws. Claude Opus 4.6 succeeded in 81 per cent of self-replication tests; GPT-5.4 reached 33 per cent. Frontier-model success on this task has risen from around 6 per cent to 81 per cent in twelve months. The experiments ran in a controlled environment with intentionally vulnerable systems - real networks have stronger defences - but the trajectory is the point.

Combined with the Beazley cyber-readiness poll that ranked construction the least prepared industry, this should make every agent rollout review one specific question: what permissions does this agent have on production systems, and how do we revoke and audit them?

Bureau Veritas - the certification and assurance body - is running a free "AI Governance & Responsible AI" webinar on Thursday 14 May, pitched at organisations that are "moving fast with AI" but "not prepared". The framing alone is a useful temperature check: certification bodies do not run governance webinars unless their client base is asking for them.