24 March 2026
Which AI model should your construction team actually use?
UK construction runs on thin margins and thick paperwork. The firms that work out how to use AI properly over the next twelve months will pull ahead. The ones that do not will keep losing commercial time to admin that a machine should be doing.
The confusion starts with the names. ChatGPT, Claude, Gemini, Perplexity, Grok. They are all different products built on frontier AI models, and each has different strengths. Think of the model as the engine, not the car. Knowing which to use where is half the battle.
None of them replace your team. They take the drudge work off the desk so your people can focus on the work that earns money and protects margin.
ChatGPT: the all-rounder
The strongest broad starting point for mixed office work. It drafts replies to RFIs, early warnings and delay letters using your own templates as examples. It summarises huge tender packs into key risks and information requests in minutes. It turns meeting notes into disciplined action trackers with owners and due dates. It also underpins Microsoft Copilot, so the same engine may already be inside your stack.
Realistic gains: three to eight hours saved per person per week on reporting, drafting and document review. If you are picking one model to start with, this is probably it.
Claude: the document specialist
The one for commercial, legal and document-heavy teams. Upload a 200-page NEC4 or JCT contract and identify onerous clauses, LDs and conditions precedent in minutes rather than days. Review subcontract inclusions and exclusions into a scope-gap summary your QS can act on. Compare two complex quotes side by side, with discrepancies in provisional sums and qualifications highlighted. Turn a full contract pack with Z-clauses into a plain-English playbook: when X happens, do Y within Z days.
If your QSs and commercial managers spend a fifth to a third of their time reading and reconciling documents, this is the one to pilot first. Fewer missed notices means less margin erosion from out-of-time claims.
Gemini: the least-change option for Google firms
If your business already lives in Gmail, Drive and Sheets, Gemini is often the easiest path because there is almost no switching cost. Search across email and Drive in one question. Turn existing Sheets KPIs into monthly board narratives. Generate formulas and pivots without writing one yourself. Ask questions of your own SOPs and lessons-learned folders.
Perplexity: the research analyst
The best first stop for anything that needs sources. Current Building Regulations requirements with links to the actual documents. Live material price movements before an estimate goes out. Client pipelines and competitor case studies before a BD meeting. Manufacturer claims verified before procurement. Use it alongside another model rather than instead of one.
Grok: the live feed
Honest assessment: useful but niche for most contractors. Its strength is real-time monitoring of news and social sentiment, which matters if you have significant planning exposure or reputational risk on public-facing schemes. For core project delivery, it is not your first choice.
Open-source models: when data cannot leave your firewall
For firms with hard confidentiality constraints, models such as Llama, Gemma and Qwen can be hosted on your own private UK server. Your sensitive project data, client lists, pricing and margins never leave your environment. This is the route for private assistants over SharePoint libraries, commercial correspondence archives and subcontractor records.
The rule that matters more than the model
Whichever engine you choose, the same principle applies: these are copilot tasks. The AI does the legwork, your people make the decisions. No automatic issue of instructions, contractual notices or compliance decisions without human review.
Start with one team, one model and two or three workflows that visibly eat hours. Prove it, then expand. That is how adoption sticks, and it is how the savings turn up in the accounts rather than in a slide deck.