AI on the board agenda: why construction needs digital evolution, not another transformation programme

Why AI is a board matter, not an IT project

For twenty years, 'digital' in construction could be delegated. A CIO ran the ERP upgrade, a BIM manager ran the modelling standards, and the board saw a budget line and a quarterly update. AI breaks that model for three reasons, and none of them is technological.

First, accountability is now regulated. The EU AI Act entered into force on 1 August 2024 and applies in stages: prohibitions and the AI-literacy duty of Article 4 since 2 February 2025, obligations for general-purpose AI models since 2 August 2025, and the bulk of the high-risk regime from 2 August 2026. Article 4 is the provision most boards have not noticed: providers and deployers - any company whose staff use AI systems at work - must ensure a sufficient level of AI literacy in the people operating them. That is a duty of the employer, and duties of the employer end up on the board's desk.

Second, the risk is strategic, not operational. An ERP that fails costs you a quarter. A competitor that learns to price, tender and manage product data with AI while you pilot chatbots costs you a market position. Construction has spent decades near the bottom of the productivity league - McKinsey's analysis of the sector found productivity growth of around one percent a year over two decades - which is precisely why the upside of getting this right is larger here than in industries that already digitalised.

Third, AI changes how your company is seen, not just how it works. Your next client's first question increasingly goes to an AI assistant, not a search box - and the assistant answers with a synthesis in which your company either appears as a cited source or does not appear at all. That is a question of market presence, and market presence is a board matter.

From adoption to autonomy: transformation has an end date, evolution has a budget line

In the technology roadmap series I have published since 2025, and tested against how the market actually moved, one arc holds across every cluster: the last two decades of digitalisation were about adoption - moving processes, data and workflows online. The stage now opening is about autonomy - systems that self-regulate, self-optimise and increasingly self-protect, running with progressively less human oversight. That arc does not pause politely at the construction industry's door.

This is why the vocabulary matters more than it seems. A transformation programme is a project: a business case, a steering committee, a go-live, a closing report. It is the right shape for replacing a system - and the wrong shape for a technology improving on a cycle measured in months, where a capability 'delivered' in one quarter is stale two quarters later. Digital evolution is the alternative posture: the organisation assumes continuous change and builds the standing capabilities that make each new wave adoptable at low cost. Three of them, each one a thing a board can actually inspect:

• Data it can trust - and controls. Structured, owned, machine-readable data about products, projects and processes, with explicit rules for what may leave the company and what stays sovereign. In construction this is concrete: BIM object data, classification, and the product records that the Digital Product Passport will soon make a condition of market access anyway.
• People who can use it. Not a data-science silo - AI-literate engineers, estimators, product managers and salespeople who know what the tools can and cannot do, and where human judgement still beats the machine. This is also what Article 4 of the AI Act now expects of you.
• Governance that keeps pace. A named owner for AI, a register of where it is used, auditable decisions, and a review rhythm - so adoption is fast because it is controlled, not slow because nobody dares to decide.

The financial consequence: stop asking 'what does the AI project cost and when does it end?' and start asking 'what does the AI capability cost per year, and what did it return this year?'. A board already governs financial control, quality and safety this way. AI joins that list; it does not join the project portfolio.

AI governance: from ethics slideware to operational law

Most companies' AI governance today is a principles document - well-meant, unenforceable, and invisible to the systems it is supposed to govern. That era is closing. Regulation and liability are converging on what I call operational law: constraints that live inside the AI lifecycle itself - logged, auditable, explainable - rather than beside it in a PDF. The practical translation for a board: treat AI governance as risk management, not moral guidance.

Concretely, that means three capabilities the board should ask to see evidence of. An AI register: which systems are in use, by whom, on what data - there are always more than the board thinks. Auditability: for decisions that matter (pricing, hiring, safety), the ability to trace why the system produced the outcome it did, because the company deploying AI without that traceability carries the regulatory penalties, the litigation and the brand damage when an outcome is challenged. And classification against the [AI Act's](https://eur-lex.europa.eu/eli/reg/2024/1689/oj) risk categories before the high-risk regime bites - a construction business using AI in recruitment, worker monitoring or safety-critical contexts may be closer to the high-risk perimeter than it assumes.

Boards that put these controls in early convert them from cost into speed: the company with a clean register and auditable pipelines is the one that can say yes to the next AI use case in a week, while its competitors convene a task force.

The construction specifics: your data decides what AI can do for you

Generic AI advice fails in construction for a specific reason: the sector's value is locked in unstructured artefacts - drawings, PDFs, spreadsheets, e-mail threads, tribal knowledge on site. A language model summarising a messy folder produces a fluent summary of a mess. The companies getting real returns are those that put structure under the AI first.

For manufacturers of construction products, the good news is that the structuring work is already mandated. The revised Construction Products Regulation makes machine-readable product data - the Digital Product Passport - part of CE marking as product families migrate to the new rules, with the first families already queued for standardisation. The same structured product data that satisfies the regulator is what makes your catalogue legible to AI assistants, configurators and your clients' BIM workflows. One investment, three returns - but only if the board treats product data as an asset with an owner, not as documentation.

For contractors and developers, the equivalent foundation is openBIM: standardised, exchangeable project data instead of file archaeology. And for both, the winning posture is not building a walled platform but joining the ecosystem: the value chains now forming - manufacturer data flowing into design tools, passports flowing into building logbooks, project data flowing between partners - reward the companies whose data can travel with privacy and sovereignty guardrails intact. Either way the sequence is the same: data first, tools second, headlines last. Boards that fund the unglamorous data layer buy themselves options on every AI wave that follows; boards that buy tools first tend to buy them again.

Your buyers already ask AI about you

A decade ago, a specifier or investor with a question about your products googled it and browsed the results. Today an increasing share of those questions go to an AI assistant, which reads the sources it trusts and answers directly. The consequence for visibility is brutal and simple: a synthesised answer has far fewer slots than a results page. You are either part of the answer, with your expertise cited, or you are invisible at the exact moment a buying decision forms.

What earns a place in those answers is not advertising spend. Engines - search and generative alike - favour sources that demonstrate verifiable expertise: named authors with real credentials, claims with dates and references, structured data that machines can check. In other words, the visibility strategy and the credibility strategy have converged. For a board this reframes marketing: publishing rigorous, attributable expert content is no longer brand polish, it is distribution infrastructure.

A practical test any director can run before the next meeting: ask an AI assistant the three questions your ideal client would ask - about your product category, the regulation hitting it, the standards behind it. Note who is cited. If your competitors' experts appear and yours do not, you have found a strategic gap that no trade-fair stand will close.

Five questions for your next board meeting

1. Who owns AI here? One accountable executive - with a mandate covering data, use cases and compliance - or a diffusion of pilots nobody can list?
2. Are we legally ready? Can we show, today, how we meet the AI-literacy duty in force since February 2025, and do we know which of our uses fall into the AI Act's risk categories before the high-risk regime bites?
3. Is our data an asset or a liability? Do our product and project data exist in a structured, machine-readable form that AI - and regulators, via the DPP - can consume, with clear rules on what may leave the company?
4. Where do we appear in AI answers? When assistants answer our clients' questions, are we a cited source? Who on the management team is accountable for that presence?
5. Where does human judgement stay in charge? As tasks move to machines, which decisions do we deliberately keep with people - and are we amplifying our people's judgement with AI, or quietly replacing it?

None of these questions requires a director to understand transformer architectures. All of them require the board to treat AI the way it treats any other fiduciary matter: named ownership, measurable exposure, honest reporting. Digital leadership in the years to 2030 will be measured not by how much technology a company adopts, but by the clarity of purpose with which it aligns people and machines.

FAQ