Construction’s AI Problem Isn’t What You Think It Is

Nearly every session circled back to data foundations. Not “what can AI do for us?” but “we still don’t have our data in order.”

Fragmented systems. Siloed information. Poor data quality. Inconsistent governance.

The most revealing session was a roundtable on data silos. Participants were frank about root causes: power dynamics, misaligned incentives, and organisational structures where information control remains valuable. As one attendee observed: silos persist because data remains a source of power and influence.

No business model discussion. How will AI reshape project economics? What happens to fee structures when design tasks that take weeks can be done in hours? Other industries are already answering these questions. Construction’s silence suggests we’re thinking about AI as something that slots into existing workflows rather than something that might demand entirely new ones.

No client perspective. The conversation was contractor- and consultant-centric. If clients don’t demand AI-enabled delivery, where’s the incentive?

Limited international thinking. For a European summit, the conversation was remarkably UK-centric.

Construction generates vast amounts of unstructured data — specifications, method statements, risk assessments, inspection reports, meeting notes, contractual correspondence.

Historically, this has been invisible to analysis. Large language models changed this overnight.

The implications are only starting to become clear.

The race isn’t to build the best AI model — the models are commoditised. The race is to build the best data estate. Construction firms, by nature of their work, already have vast estates.

They just don’t know it yet.

1. Construction’s AI problem is not an AI problem
   The challenges are downstream consequences of older failures: data infrastructure, digital capabilities, commercial relationships, leadership. AI can’t solve these. It can only amplify them.

2. The biggest blocker is structural, not technical
   The construction industry still views digital as a support function rather than strategic capability. This isn’t a failure of understanding — it’s a reflection of how the industry is structured, incentivised, and measured. The mid-tier gets it. The challenge is creating the conditions where acting on that understanding makes commercial sense.

3. Digital product design is a profession
   Construction treats software development as something engineers can do on the side. The results are predictable: poor UX, low adoption, abandoned tools. AI won’t fix bad product design. It will make the consequences more visible and more costly.

4. The technology is already here
   The AI models work. The infrastructure exists. The use cases are clear. What’s missing is organisational will and professional digital practice.

5. The variance is the opportunity
   Some organisations are genuinely advanced. Others are still arguing about whether to use a common data environment. For those who can bridge this gap, the next five years will be transformative.

Construction is three to five years behind comparable industries in AI maturity, and the gap is widening.

The technology is not the constraint. What’s missing is organisational maturity.

For those working at the intersection of digital product design and civil engineering, this is simultaneously frustrating and exhilarating. The barriers are addressable. The potential impact is immense.

Here’s the thing, though: the same AI capabilities that expose these gaps can close them. Legacy formats, fragmented data, inconsistent taxonomies — these are exactly the kinds of problems that modern language models and retrieval systems were born to solve. AI doesn’t need your data to be perfect. It can ingest a decade of PDFs, reconcile inconsistent naming conventions, bridge systems that were never designed to talk to each other, and start delivering value from day one.

The tools available right now are not just good — they’re extraordinary, and they’re improving at a pace that makes even cautious optimism feel conservative. Construction doesn’t need a five-year data transformation programme before it can start. It needs to start, and let AI do the heavy lifting it’s already capable of. The firms that grasp this — that lean in fully rather than waiting for perfect conditions — won’t just catch up with other industries. They’ll leapfrog them. Construction’s messy, rich, real-world data is not the obstacle. It’s the advantage.