The Use-Case Vacuum: 39% of Construction Firms Can't Name Where AI Goes

The UK's official statistics just exposed construction's real AI problem, and it isn't the one the vendors are selling against.

The Office for National Statistics, in its July 2026 analysis of AI use across UK businesses, put construction at 13% adoption against 58% in information and communications. The widest sector gap in the release, and the number everyone quotes [1]. But buried in the same data is the figure that actually matters. In the ONS management survey, the most commonly reported barrier to adoption, cited by 39% of firms, was difficulty identifying activities or business use cases [2]. Cost came in at 21%. Skills at 16% [2].

Read that again. The industry's biggest obstacle to AI isn't money or talent. It's that most firms cannot name a single place in their business where the technology belongs.

The vendor story assumes a buyer who exists

Every ConTech pitch is built on an implicit assumption: the buyer knows their problems and is shopping for solutions. Products are demoed against pain points, ROI cases are modelled on efficiency gains, procurement is framed as a choice between options. The ONS data says that for most of the industry, the conversation never gets that far. There is no shortlist, because there is no named problem.

That reframes what "adoption barrier" means. Cost and skills are constraints on executing a decision. The 39% is upstream of all of that: it's the decision itself that's missing. You can subsidise software and train staff all you like, but nobody buys a tool for a process they've never written down.

What unmapped process looks like

On any live project, the work AI could realistically touch is not hiding. It repeats weekly and follows a format. Payment applications assembled afresh from cost reports every month. Site diaries written from notebook scraps at 6pm. Submittal registers chased by email. Tender queries typed out one at a time from drawing PDFs. Each of these has an owner, a deadline, a format and a history of small failures. Each one is a candidate use case with its name already on it, waiting for someone to spend an hour asking what data it consumes and where the bottleneck sits.

That hour is what's missing. A firm that cannot name its use cases is a firm that has never drawn its own processes end to end. The 39% isn't measuring a technology gap. It's measuring an unmapped business.

Why the finance office moves first

Intuit shipped a construction edition of its ERP in February, and the instructive part is not the feature list. It's that the buyer never has to identify a use case at all. Phase-level job costing, progress invoicing, overrun alerts: the product arrives with the use cases pre-named [3]. The vendor did the mapping and priced it into the box. Set that against a point tool that requires the contractor to first articulate where it fits, and you can see why finance-led packages spread through the mid-market while cleverer site-side tools stall in pilots.

There's a second-order effect worth naming. When buyers can't name their own use cases, whoever names the use cases for them owns the purchase decision. Right now that's the accountant. Advisory and accounting firms have quietly become the kingmakers of construction AI for smaller contractors, because the software conversation enters through the books. A recommendation from the firm that runs your month-end carries more weight than any conference booth. Process knowledge is migrating to whoever ships first, and most of them have never set foot on site.

The fix is one page per workflow

The remedy is unglamorous and free. Pick one deliverable that repeats every week. Write down its inputs, its outputs, its format and where it stalls. Ask whether a tool could do the repetitive half. Then pick the next one. One page per workflow, a fortnight of evenings, and the firm goes from "we can't identify use cases" to a ranked list of candidates with owners attached. That list is also the only procurement shortlist worth having, because it was built from the business rather than from a demo.

This is the gap akil.build is built around: workflow blueprints and a construction knowledge graph that map deliverables, data and bottlenecks before a single tool gets shortlisted, so the buying decision starts from a named process. Whatever route a firm takes, a consultant, the accountant's recommendation, or a Saturday with a whiteboard, the principle is identical.

Name it or be named for

The 39% will not be solved by better models. It is a naming problem, and naming is done with process knowledge, which the industry holds in abundance and has never written down. Firms that map their own workflows will buy deliberately and keep the terms of the decision. The rest will have their workflows mapped for them, by vendors and advisers. The vacuum gets filled either way.

Sources

[1] ONS. "Artificial intelligence in UK businesses: 2023 to 2026." July 2026. https://www.ons.gov.uk/businessindustryandtrade/business/businessservices/articles/artificialintelligenceinukbusinesses/2023to2026

[2] ONS. "Management practices and the adoption of technology and artificial intelligence in UK firms: 2023." March 2025. https://www.ons.gov.uk/economy/economicoutputandproductivity/productivitymeasures/articles/managementpracticesandtheadoptionoftechnologyandartificialintelligenceinukfirms2023/2025-03-24

[3] Intuit. "Intuit Launches New AI-Powered Construction Edition for Intuit Enterprise Suite." February 2026. https://investors.intuit.com/news-events/press-releases/detail/1302/intuit-launches-new-ai-powered-construction-edition-for-intuit-enterprise-suite