Engineering and construction firms have a particular relationship with risk. Their work is signed, stamped and built, and errors can stand for decades. That makes the sector a useful test of where AI earns its place in a business and where it does not. Over the past two years, the large international consultancies have published a good deal about their experience. Smaller Swiss engineering offices, planners and contractors can learn from it without repeating the expensive parts.

The prototype year, and the scaling problem

Speaking on an industry podcast summarised by Roads & Bridges in December 2025, two WSP specialists, Chris Harman and Jay Wratten, described 2025 as a year in which engineering firms built small AI tools to automate tasks and improve workflows. Those prototypes worked in limited settings but were not ready for company-wide use. Wratten predicted that 2026 would be about deciding which ideas could scale, pointing to the large gap between an AI-generated workflow that suits one person and a tool thousands of employees can rely on.

That gap exists in a ten-person office too. A spreadsheet macro or chatbot prompt that one engineer finds useful is not the same as a process the whole team can trust, document and hand over when that engineer is on holiday.

Where AI is proving useful

The examples that recur across the sector fall into three groups.

Early design options. In a January 2025 insight piece, WSP described generative AI tools that can evaluate hundreds of structural options at concept stage, comparing carbon, cost and constructability, work that previously took months of detailed engineering. The ASCE's Civil Engineering Source reported in February 2026 that Arup uses a genetic algorithm to narrow billions of net-zero design options to a few thousand, which are then ranked by cost and efficiency.

Checking the built result. The same WSP article described a construction monitoring tool that compares site photographs with the engineering model and can detect deviations of a few millimetres, flagging clearance conflicts before later trades are affected.

Keeping knowledge in the firm. ASCE reported that Thornton Tomasetti used an in-house language model, with the family's approval, to build a searchable record of a veteran engineer's knowledge from seven years of his emails. Its chief technology officer was candid that extracting project intelligence for younger engineers is a goal the firm has not yet reached.

For a smaller firm, the third group may matter most. Many Swiss engineering offices depend on a handful of senior people whose judgement is not written down anywhere.

Why the engineer's role shifts towards verification

The same Roads & Bridges piece carried a prediction from Wratten worth taking seriously: as AI handles more of the production work, the engineer's job tilts towards quality assurance, checking that what has been produced is safe and complies with code. WSP's own framing is similar. Its building design article stressed that the output depends on the quality of the underlying project data and on the experience of the engineers guiding the tool, and it named hiring and retaining engineers as the firm's most important AI investment.

WSP's chief executive, Alexandre L'Heureux, made a related point when he spoke to The Canadian Press in May 2026. He described AI as a tool to augment the firm's work for clients, not a threat to it, and placed it in the context of an industry-wide labour shortage, with the global pool of engineers growing by only about one per cent a year. Swiss firms will recognise that shortage.

There is also a commercial point. Wratten also suggested that firms may move away from pure time-and-materials billing as AI shortens timelines. If your fees are built on hours, faster production is not automatically a gain.

A checklist for a Swiss engineering or construction SME

  • Start with internal knowledge, not design. Make your past reports, specifications, standards notes and lessons learned searchable before asking AI to produce anything new.
  • Use AI for options, not for the stamp. Let it generate and compare variants. Keep calculation, verification and sign-off with a qualified engineer.
  • Record what was AI-assisted. Note in project files which outputs were generated or checked with AI, so a reviewer knows where to look harder.
  • Check your contracts and insurance. Professional liability does not move to a software vendor. Confirm what your client contracts and indemnity cover say about tools.
  • Protect project data. Client drawings and site data should not go into public tools. Decide where data is hosted before any pilot begins.
  • Think about fees early. If a task takes a third of the time, decide whether you bill for outcome, time or a technology line.
  • Pick one bottleneck. Tender document review, site report drafting or deficiency lists are bounded, repetitive and easy to measure.

The bottom line

The large consultancies have shown that AI can widen the set of options an engineer considers and catch errors earlier on site. They have also shown that prototypes are easy and dependable tools are hard, and that responsibility stays firmly with the professional who signs. For a smaller firm, the sensible path is to capture what your senior people know, automate one bounded task, and keep the engineer's judgement at the centre.

Curious where AI could relieve your team's capacity? The free AI Readiness Audit maps your recurring tasks and ranks what is worth automating first.

Sources

  • Roads & Bridges, "Predictions for 2026", 29 December 2025. link
  • WSP, "Generative Artificial Intelligence and Building Design: Solving the Unsolvable", 21 January 2025. link
  • ASCE Civil Engineering Source, "Engineering firms could add AI roles to organizational charts sooner than you think", 11 February 2026. link
  • BNN Bloomberg (The Canadian Press), "AI a 'great thing' for engineering amid labour shortage, says WSP Global CEO", 7 May 2026. link