Kennisbank

AI governance in the installation industry: where do you start

An industry that runs on planning and execution

The installation industry consists of a combination that few other sectors have: a lot of physical work in the field, alongside a back office that calculates, plans and administers. These two worlds use AI in a different way and to a different degree. The field service works with fault diagnosis, technical manuals and sometimes image recognition for inspection. The back office works with quotes, calculations, planning of technicians and communication with clients and subcontractors. Where in one department AI is mainly a tool for a task, in the other department it is a link in a chain of decisions that eventually ends up in an invoice or a contract.

This ratio — more executional work on one side, more administrative and commercial work on the other — determines where the risk lies. An AI tool that helps a technician interpret a fault code has a different impact than a tool that automatically generates a price quote for a customer. Anyone setting up AI governance without that distinction treats both cases the same and thereby misses where attention is most needed.

Why the IT list is more often empty here than elsewhere

Installation companies are often organisations with a compact IT department and a large field service that works independently. Technicians, calculators and planners search on their own for tools that make their work easier, without any purchasing process preceding it. A free chatbot for drafting a quote, an app that analyses photos of an installation, a tool that compiles material lists: these are choices made on the shop floor, not at head office.

This means that the list IT can provide is, in this industry, more often an underestimate than in sectors with a stronger central IT function. Anyone who uses only that list as the basis for governance builds on a foundation that is not complete. The only way to correct that is to ask the people who do the work — and that conversation only yields something if there is no sanction attached to it. A technician who is afraid of a note in his file will not mention the tool he uses to write a quote faster.

Role and risk as an organising principle

An AI scan that does justice to this industry does not organise by department but by role and risk level. A tool that summarises technical documentation for a technician carries a different risk than a tool that gives a price indication to a customer without human oversight. A tool that optimises planning for the organisation itself is different from a tool that processes customer personal data in a CRM integration. This classification makes visible which applications need attention and which can continue to exist side by side without further measures.

This way of looking returns in other sectors with a comparable spread of work, as can be seen in what is described about AI governance in the transport sector, where planning and execution are also separated. The comparison with AI governance in the manufacturing industry is also relevant, because there too a production environment comes together with an administrative layer that carries different risks. And anyone looking at AI governance in wholesale sees a similar pattern around quotes, pricing and the question of who bears responsibility for that.

Connecting to what is already there

Installation companies often already work with quality systems, safety protocols and certification requirements that arise from the nature of the work. AI governance does not need to be built alongside that structure, but can be incorporated into it. A risk classification for AI applications connects to existing risk assessments; responsibility for an AI tool in the quotation process fits with the existing responsibility for pricing. The question is not whether a new system is needed, but whether the existing system has been extended with the question: what role does AI play here, and who is responsible for it.

The current regulation around AI systems, including classification and obligations, is changing and is tracked elsewhere. What is described on this page is the mechanism: taking stock of what is running, classifying by role and risk, and fitting that into the governance that already exists.

From inventory to insight into the work

An inventory of AI use in the installation industry shows which tools exist and what risk they carry. That is a different question from the question of which part of the work itself — the calculating, the planning, the drafting of quotes, the diagnosing of faults — is suitable to be taken over by AI. That question is answered in the work scan from FTE TO AI, which calculates per task which part of the work can be taken over and which part remains with the human. Where this page is about grip on what is already running, that scan is about the direction in which the work is moving.