AI in construction: 67% want to invest, 8% have — the brake is trust (and EdilIA is coming)

Construction is the sector that wants to move and can't. It's an almost paradoxical picture: 67% of Italian building firms say they are willing to invest in AI tools and 51% already consider digital and AI indispensable to stay competitive — yet real adoption is the lowest of any sector, with barely 9.7% of firms even considering it at European level and just 8% of Italian SMEs running a structured project, against 71% of large companies. The appetite is there. The adoption isn't.

The gap, in construction, doesn't read the way it does elsewhere. It isn't cost, as on the factory floor; it isn't skills, as in retail; it isn't the shortage of staff, as in logistics. It's something harder to buy with a grant: it's trust. And for a trust-blocked sector, the right reading turns every demo on its head — trust comes before the tool.

The real obstacle isn't the price: it's believing the result

When you ask a building firm why it hasn't adopted AI, the first answer isn't "it costs too much". It's scepticism about the reliability and accuracy of what the AI produces: the person who answers to a client, a final inspection or a soprintendenza (the heritage-protection authority) won't trust an output they can't explain. Right behind it comes the difficulty of integration with project infrastructure that is often still paper-based or fragmented, and finally the learning curve.

It's a different barrier from every other sector: the accountant has the tool but underuses it, the factory is blocked by cost, retail by skills, logistics by staffing. Construction is blocked by trust — which makes it closer to a change-management sale than a software sale. You don't win it with one more licence: you win it by proving, on a low-risk case, that the result holds up.

The State is digitalising the permit: why the urgency is already real

There's a reason this topic can't sit at the back of the queue: the State is automating the permit itself, and the deadline doesn't depend on any single firm's willingness. EdilIA (the national AI system for building permits), introduced by decree (DPCM) in 2026, aims at an automatic response in 180 seconds to building-authorisation requests via SPID (Italy's public digital identity login), with the goal of zeroing review (istruttoria) times and reassembling the fragmentation between comuni (municipalities), sportelli unici (one-stop building counters) and soprintendenze. In parallel, BIM (Building Information Modelling, the shared digital model of a building) is mandatory for public tenders above 2 million euro from 1 January 2025.

The reading for anyone who builds or designs is sharp: while the State digitalises the entry point of the application, you'll need to arrive at the door with a submission already prepared — modelled in BIM, checked against the rules, clean — before it meets EdilIA. It's a wedge built by the legislator, not a threat: it creates real urgency, independent of any sales effort, and shifts the advantage to whoever gets ready in time. It's the same regulatory ground our building-permit tool works on — we know it from the inside.

Where to start: the low-risk pilot that builds trust

If the barrier is trust, the mistake is to start from the most spectacular use case — automatic permit verification on a complete BIM model. It's the most visible, but also the one where an error costs the most and the distrust is highest. The sensible path is the opposite: you begin with the case that carries the lowest risk and the most legible before/after, you measure it, and only then — with the proof in hand — do you climb to the cases that touch compliance.

The four use cases, from the most controllable to the most regulated

The sector evidence points to four concrete use cases for a construction SME. They're worth reading in the right order — not by technological maturity, but by how easy it is to trust the result:

• Forecasting delays and variations — a predictive alert in place of the hand-updated Gantt chart and reactive variation orders. It's the natural candidate for a first pilot: the output is an early warning, not a binding decision, so an error costs little and trust builds early.
• Digitalising the site — the "Fascicolo Digitale delle Costruzioni" (Digital Building Record) — a structured, queryable dossier in place of the paper archive per property. It's a building-level "digital twin" that moves documentation from the folder to a searchable record: immediate value, low risk, and the foundation the more advanced cases rest on.
• Automatic verification of permits and compliance on BIM/IFC models — the "ePermit" / Digital Building Permit approach: the AI, with the rules and the model, checks a submission against the regulations before filing, cutting review iterations. It's the highest-value case ahead of EdilIA, but also the one that touches the filing: tackle it after building trust elsewhere.
• Predictive maintenance on machinery and plant — acting on degradation signals in place of calendar-based maintenance. The same downtime-reduction mechanism documented in manufacturing applies to site machines and plant.

The criterion isn't "which is the most advanced" but "which can I control and believe with the least risk": in a sector where the block is trust, the order in which you adopt the cases matters as much as the cases themselves.

And compliance? Here it's part of the product, not an add-on

In construction, compliance isn't a side topic: it's the ground itself. A system that checks a filing against the regulations, or that touches the authorisation to build, has to be handled rigorously under the EU AI Act — with an impact assessment where needed, clarity on where the vendor processes the project data, and above all a documented human oversight: the AI checks and flags, the signature on the filing stays with the professional who takes responsibility for it. That's exactly the point that resolves the initial scepticism — not "trust the AI", but "the AI prepares, you check and sign". Our compliance overlay wires these controls to every workflow we design, so the adopted case is also defensible in front of a client or a soprintendenza.

Where to start, in practice

If you build or design and AI is in your sights, the sensible path is short and ordered — designed to win over the distrust, not to bypass it:

• Start from the lowest-risk case — forecasting delays and variations, or digitalising the building record. The output can be checked, an error doesn't block a filing, trust builds early.
• Measure the before/after — delays anticipated, review hours saved, variations avoided. A number the site understands and that proves the result holds up.
• Move up a case only afterwards — verification on a BIM model ahead of EdilIA comes when trust is already built and the method is proven, not before.
• Keep the human signature on the filing — the AI prepares and checks, the responsibility stays with the professional. That's what makes the result defensible, not just fast.

Even before choosing the case, though, it's worth knowing where you are: our AI-readiness assessment helps you understand where to start with more return and less friction, and which controls to put around the first pilot. If the theme is the compliance of what touches the filing or the project data, our compliance overlay explains how we wire the controls to every design.

We've turned the first step into a self-serve, free assessment: a few questions and an indication of where to start, with what controls around it. Take the AI-readiness assessment — then, if it makes sense, let's talk.

This article is for orientation. The figures on adoption, willingness to invest and return, and the regulatory references cited (EdilIA, introduced by decree (DPCM) in 2026, with the goal of an automatic response in 180 seconds via SPID; BIM mandatory for public tenders above 2 million euro from 1 January 2025) come from industry sources and institutional communications — chiefly Norasoft (2026) for the structured-adoption figures (8% of SMEs vs 71% of large firms; 9.7% EU-level consideration), LEVELS (2026) for willingness to invest (67%) and the perception of indispensability (51%), LavoriPubblici for EdilIA/DPCM and BibLus/ACCA for the BIM obligation — and should be read as indications of direction, not as guarantees of results: timings, thresholds and requirements must be verified against the official texts of the measures before any decision. Every tool choice and every automation that touches a building filing, project data or the authorisation to build must be assessed against the data, the controls and the context of the individual firm, with the oversight of a qualified professional.