Dijkhoff leads the way in civil engineering and puts AI to work on BPKV tenders, specifications and CROW 400

Where most civil engineering firms are still waiting, Dijkhoff B.V. put its whole company through a tailored training day: the AI training for the civil engineering sector. Management, work preparation, QHSE, HR and administration sat in one group of six and practised on their own work, such as BPKV tenders, specifications and CROW 400. Across six evaluations the recommendation score came out at 8.5/10.

The challenge

Dijkhoff started in 1914 trading bricks and coal and is now a family business of over a hundred years in ground, road and water works, with demolition, nature engineering and circular recycling alongside it. You can see that it leads on the CO2 Performance Ladder: the company sits at level 5, the highest rung, and calls that certification the natural consequence of how it works. When AI entered the craft, in BPKV texts, specifications and estimating, the reflex was no different. Everyone was already using ChatGPT, a third also Microsoft Copilot.

So the issue was not a lack of interest, but that it was happening unmanaged: everyone on their own, without a shared standard. Management therefore took the initiative itself, in a sector that consistently ranks at the bottom on digitalisation. And with the requirements you would expect from an ISO and VCA certified company: the quality and accuracy of AI output has to be verifiable before anything reaches a client (83% put that first) and information security has to hold up (67%). On top of that came the labour shortage in the sector, named by four of the six: whoever learns now where AI takes over work will be ahead later.

The approach

We started with an intake among all six participants. That intake immediately showed why company-wide was the right call: the baseline ranged from just starting out to well experienced (average 2.8 out of 5, range 1 to 5) and for five of the six this was the first AI training of their career, because nobody in the sector had offered one before. That led to one day in Heeswijk-Dinther for the whole group, from management to QHSE and work preparation, so that everyone spoke the same language about AI afterwards.

The morning covered how generative AI works, effective prompting, reliability and privacy (from hallucinations to the transparency obligation in the AI Act), assessing AI output, and the tools themselves including AI agents. After lunch it was all about Dijkhoff: BPKV tenders, breaking down specifications, CROW 400, soil reports and route optimization, ending in a workshop where everyone worked on a task of their own.

The result

The training scored a recommendation score of 8.5/10 across six evaluations, with a trainer rating of 4.8/5 and learning goals at 4.0/5. Five of the six participants also named at least one task themselves that they will use AI for.

Top of the list were weekly reports, BPKV tenders, specifications, resident letters, estimating and route optimization for transport. We also showed what is possible with CROW 400 and QHSE documentation and with handover dossiers built from project data. There is also a plan for an in-house AI knowledge base that unlocks over a hundred years of expertise, so the experience of the people about to retire does not walk out of the door with them. Dijkhoff now has a group that knows what AI can and cannot do, and a starting point for its own internal policy.