AI in the trades and construction: adoption is real, but it is concentrated in the office

Construction and the skilled trades are often described as an industry waiting for AI. The published evidence says something more specific: some firms are using it every week, most are not, and the tasks that work today are mostly administrative. This piece summarizes what the surveys show, with their limits, and what that implies for contractors and field-service operators deciding where to start.

How many firms are using it

The most-cited recent figure comes from Houzz's 2025 survey of more than 700 construction and design professionals. Overall, 34 percent of professionals said they already use AI in their businesses; in construction specifically the figure was 32 percent, rising to 64 percent among large firms ^[1]. The release does not describe the sampling method or margin of error, so treat the figures as indicative rather than representative of the whole sector ^[1].

Other published data paints a slower picture. Activant Capital's November 2025 research note cites the RICS AI in Construction 2025 survey, where about 45 percent of respondents said their organizations do not use AI, 34 percent were in early pilots, and fewer than 12 percent regularly used it in a specific process. The same note cites an MIT Sloan study reporting that 65 percent of construction leaders said their companies were not yet using AI or predictive analytics for planning or execution ^[3]. Those are second-hand citations from an investor research note, not the original studies, but they agree with Houzz on the broad shape: a minority of active users and a large group still watching.

What people actually use it for

Among construction respondents in the Houzz survey, the leading uses were administrative work (47 percent), project and client management (42 percent), sales and marketing (37 percent), and planning and design (32 percent) ^[1].

Nothing in that list involves a robot on a site. The work being automated is the paperwork around the job: documents, scheduling, estimates, follow-ups and customer communication. Activant makes the same point from the other direction, noting that experts see the largest impact in pre-construction (design, estimating and bidding) and that AI could take over the "second job" of project managers and superintendents: scheduling, reporting, compliance and other administration ^[3].

Why the back office comes first

Three reasons show up in the sources.

The data exists in documents, not in sensors. Activant notes that construction's slow digitization leaves data unstructured and processes manual, creating a paradox in which AI depends on digitization while digitization stalls without clear AI value ^[3]. Language models are comparatively good at unstructured text, so documents are where they help first.

Systems do not talk to each other. Activant cites an Autodesk finding that 51 percent of construction teams manually transfer data between apps that do not integrate ^[3]. Connecting those systems is integration work before it is AI work.

Errors are costly on site and cheap in the office. The same note observes that contractors are hesitant to adopt tools that are not nearly flawless where errors could lead to catastrophic outcomes ^[3]. A wrong draft email can be corrected; a wrong structural assumption cannot.

The skills gap

The Houzz release reports that among professionals already using AI, 92 percent of contractors and 94 percent of designers have no formal instruction in applying it to their work ^[1][2]. Among those using AI, concerns centered on reliability and accuracy of outputs, data security and privacy, and lack of training; the release gives no percentages for them ^[1]. This matters because the productivity claims in the same survey (more than three hours saved per week, over $108,000 in annual gains per business) are self-reported and calculated by Houzz from time savings and average revenue ^[1]. They are an upper-end estimate, not an audited result.

Where the evidence runs out

Field-service companies spend much of their effort on intake, scheduling and customer communication, but none of the sources above measures outcomes for that work. They report how many firms use AI and for what, not whether it improved speed, cost or quality in the trades. The Houzz time-savings figures are self-reported estimates, and the Activant material is second-hand.

The closest controlled research we reviewed is not about the trades at all. An NBER study of 5,179 customer support agents found that access to a generative AI assistant raised issues resolved per hour by about 14 percent on average, and by about 34 percent for novice and low-skilled workers, with minimal effect on experienced workers ^[4]. That is a 2023 call-center study. It says nothing direct about contractors, dispatchers or estimators, and we cite it only as an example of how such effects can be measured, which is a reason to measure your own.

What this means for contractors and field-service operators

Start with one administrative workflow that is repetitive, document-heavy and reversible: quote follow-ups, job summaries, or intake notes. Have a person review the output, measure time saved on that single task for a month, and invest in basic training before buying more tools. Defer anything that touches safety-critical judgment on site until you have a track record in the office.