AI in the Trades: What Contractors Actually Adopted

The short version

Two rates for contractor use of artificial intelligence (AI) published in 2026 are under 10 percent and 81.5 percent, and both are defensible, because the organizations behind them asked different questions of different populations. The Census Bureau asked employer businesses whether they used AI in any business function during the previous two weeks, collecting the answers between November 17, 2025 and February 8, 2026, and the construction sector came in under 10 percent. Jobber asked 1,050 home service business owners whether they use AI, fielding that survey in December 2025, and HVAC came in highest of any trade at 81.5 percent. Neither one is the adoption rate, and treating either one as the adoption rate is what costs money.

A shop that buys seats nobody logs into pays a monthly subscription it has to recover in its labor rate, and the homeowner pays that rate. A shop that dismisses the category because the Census Bureau puts construction under 10 percent keeps sending after-hours calls to voicemail, and a customer who cannot reach anyone calls the next contractor on the list.

What got adopted is narrower than any headline rate suggests. Every publisher that broke adoption down by business function put the heaviest use in office work: quoting and invoicing for Jobber, sales and marketing for Houzz, office applications for AGC and Sage. Census Bureau researchers report that 57 percent of AI users apply it in three or fewer business functions, so even where adoption happened it is usually not broad. None of the function breakdowns in this post has a line item for diagnosing equipment at the unit.

Eight numbers published inside nine months

Eight numbers for contractor AI use appeared between December 2025 and the summer of 2026, and their publishers did not all measure adoption. The table below gives who published each one, who answered it, and what each publisher measured, because who answered and what got measured explain differences the headline numbers do not show.

One definition first, because the sources below use both forms and they are not the same claim. A percentage point is the arithmetic difference between two percentages. Going from 44 percent to 61 percent is a rise of 17 percentage points, and it is also a rise of about 39 percent relative to where it started, since 61 divided by 44 is about 1.39.

Who published it	Who answered	When	What was measured	The result
The Census Bureau, Business Trends and Outlook Survey AI supplement, as read by the Federal Reserve Bank of Minneapolis	Construction businesses inside a national sample of employer businesses	Collected November 17, 2025 to February 8, 2026; supplement reported April 2026	Whether the business used AI in any of its business functions in the last two weeks	Under 10 percent
ServiceTitan, 2026 State of AI in the Trades	1,032 commercial and residential contractors across seven trades, including company leadership, office staff and field technicians	Fielded October 23 to November 12, 2025; published December 16, 2025	Whether AI is embedded in operations today, or being experimented with	12 percent embedded, 34 percent experimenting
ServiceTitan, 2026 Residential State of the Trades	1,000 residential owners, executives, general managers and directors, polled by Thrive Analytics	Released April 7, 2026	The share of contractors currently using AI	About 25 percent
AGC and Sage, 2026 Construction Hiring and Business Outlook	951 respondents from construction firms across 49 states and the District of Columbia	Fielded November 4 to December 15, 2025	Whether the firm initiated or increased its use of AI in the past year	61 percent, up from 44 percent
Houzz, 2026 State of AI in Construction and Design	601 U.S. construction and design businesses	Fielded June and July 2026	The share using AI for everyday business tasks	52 percent of construction firms
Jobber, 2026 Home Service Trends Report	1,050 home service business owners across eight trades	Fielded December 2025 through Conjointly	The share using AI in day-to-day operations	52 percent overall, 81.5 percent in HVAC
Dodge Construction Network with CMiC	235 general and trade contractors	Fielded September and October 2025	Whether the contractor expects AI to have a meaningful impact on the construction industry, and whether it has adapted its workflows	87 percent expect a meaningful impact, 19 percent have adapted workflows
ServiceTitan, 2026 Commercial Specialty Contractor Industry Report	More than 1,000 construction leaders	Published 2026	The share reporting measurable business impact from AI	38 percent, up from 17 percent in 2025

The spread across those eight rows is at least eight to one, since the low end is published only as under 10 percent, and the dated reports among the eight appeared inside about nine months.

Three of the rows come from one company, and you can see the point from two of them alone. ServiceTitan reports that “only about 25% of contractors are currently using AI” in its residential report, and in its trades AI report that “Only 12% have embedded AI into their operations today, and 34% are actively experimenting.” That comparison shows the publisher’s name is not what sets the rate. The question moves it, and so does who answered, since one of those reports polled 1,000 residential owners and executives and the other polled 1,032 residential and commercial contractors including field technicians.

Houzz, a home renovation and design platform, published two rates from one survey of 601 businesses, and both get quoted as though they cover the whole industry. Houzz reports that “Forty-one percent of firms now use AI for business tasks, up 7 percentage points from a year ago” across construction and design together, and that “More than half of construction firms (52%) now use AI tools for everyday business tasks, up 20 percentage points from a year ago.” Both rates are real. They cover different groups of respondents.

The Jobber rate needs the same care. Jobber surveyed 1,050 home service business owners across cleaning, HVAC, roofing, plumbing, electrical, lawn care, tree care and general contracting, so the HVAC owners are a fraction of that 1,050 rather than all of it, and Jobber does not publish how large that fraction is. The 81.5 percent is the highest adoption rate in this post and it depends on a subsample of unpublished size. That is a reason to quote it with its context rather than a reason to throw it out.

Why the rates disagree

The spread depends on four choices the publisher makes, and none of those choices is about how many contractors adopted anything.

Who answered. The Census Bureau samples employer businesses across the economy, excluding farms, and most American firms are very small. A Federal Reserve Board policy analyst makes the point directly: “The vast majority of firms in the U.S. are small. For additional context, about 57 percent of firms have fewer than five employees.” The same analyst adds that the Business Trends and Outlook Survey “generally mirrors the firm population distribution.”

Firm size moves this measurement hard. The Census Bureau reports that “37% of firms with at least 250 employees reported using AI in their business operations,” against “Less than 20% of firms with four or fewer employees.” A survey recruited by a software company reaches a different population from every business holding a contractor license, though not necessarily a population limited to that company’s own customers: Jobber fielded its survey through Conjointly, a survey research firm, and ServiceTitan fielded its trades survey through Thrive Analytics, a market research firm.

What the question counted. The cleanest evidence here comes from one organization measuring one population twice. The Census Bureau changed the wording of its AI question on November 17, 2025, from whether the business used AI “in producing goods or services” to whether it used AI “in any of its business functions.” Nothing else about the question changed.

A Federal Reserve Board policy analyst puts the effect this way: “The jump in reported adoption between the new and legacy series ranged from 47 percent (10.6 percentage points) for professional services on the low end to 159 percent (7.5 percentage points) for manufacturing on the high end.” The Census Bureau started a new time series rather than continuing the old one.

That change is this post’s whole argument in one example. With the narrower wording, the Census Bureau asked whether AI is in the work the business sells. With the broader wording, it asks whether AI is anywhere in the business, which takes in the office manager using a chatbot to write a follow-up email. The same analyst describes the original wording as the conservative one, saying the survey “was originally designed to capture significant AI usage through more conservative question framing that focused on usage in producing goods and services.”

Whether the summary matches the question. The Associated General Contractors of America (AGC) and Sage report that “Sixty-one percent of respondents say their firms are using artificial intelligence or plan to increase investment in it, up from 44 percent last year.” Their published national survey asked something narrower than that sentence describes. Question 15 reads “Has your firm initiated or increased its use of artificial intelligence (AI) in the past year for any of the following?” and offers seven functions plus “None of the above,” which 39 percent of the 857 firms answering chose. The 61 percent is the rest of them, so it counts firms that started or expanded AI use in at least one function during the past year rather than firms that only plan to spend money on it. Repeat the headline sentence and you carry along a description of the question that the questionnaire does not support.

Who is asking. A Federal Reserve Board policy analyst names this one: “Recently, though, firm representatives, especially senior leaders, may face pressure to report AI usage as an efficiency initiative. This could put upward pressure on estimates that target corporate leaders.” Nobody in that description is lying. An owner who switched on one AI feature in software the shop already pays for, asked by that software company whether the business uses AI, will reasonably answer yes.

What contractors actually put into daily use

Ask what AI is doing rather than how many shops have it, and the picture gets more specific. Three reports give a breakdown by business function, and each one’s leading category is office or business work rather than work at the equipment.

Jobber reports that among home service owners using AI, “54% use it for quoting, 52% for invoicing, and 51% for business writing” such as emails, proposals and job descriptions. Below those come marketing content at 46 percent, customer communication at 35 percent, analyzing customer feedback at 33 percent, and scheduling and dispatch at 29 percent, where dispatch means assigning technicians to calls and routing them through the day.
Houzz reports that AI “is being used for sales and marketing leads (64%), followed by planning and design (61%), project and client management (59%), administrative work (52%) and business operations (45%).”
AGC and Sage report that “Forty-five percent of firms deploy AI for office and administrative functions,” with “23 percent” using it for estimating work, “20 percent” applying it to design or preconstruction activities, and “16 percent” using it for recruiting, training, or other human resources functions.

Census Bureau researchers report the same leading function across the broader business population, working from the 2026 AI supplement to the Business Trends and Outlook Survey. Adopting firms use AI, in those researchers’ words, “most commonly Sales and Marketing (52%), Strategy and Business Development (45%), and IT (41%).” The same researchers report that “57% of users integrate AI in three or fewer business functions,” which means adoption is narrow inside the firms that have it.

None of those lists has a line item for a technician diagnosing equipment on a service call. The closest is AGC and Sage’s “Onsite construction activity, monitoring, documentation,” selected by 13 percent of firms, and that option covers monitoring and documentation rather than diagnosis.

Some of what got adopted arrived inside software the shop already ran. The Census Bureau’s cognitive testing on its own AI question found respondents who first answered no and then turned out to be using AI “provided as part of software workflows,” which is adoption nobody shopped for. It is not the whole story. Quoting and invoicing lead Jobber’s list and not the other two, and the AI features bolted onto these platforms are not all free, since Jobber says its Receptionist, its phone answering feature, “is available as an add-on to any plan, and is included at no additional cost with the Plus plan.”

Jobber describes its Receptionist as a feature that “answers calls and texts 24/7,” that “can capture job requests and book visits during the call,” and that “creates a task in your calendar to follow up, along with a summary of the conversation” when somebody leaves a message. Housecall Pro describes its customer service representative AI in similar terms, as a feature that “Answers calls and books jobs 24/7 so you never miss an opportunity.” Both companies solved a problem every shop recognizes, which is that the phone rings while everybody who could answer it is already on a call.

Answering the phone is the clearest product story in this trade, and no survey in this post puts a number on it. Jobber, Houzz, and AGC and Sage all published function breakdowns, and none of the three carries a line for answering the phone, so the vendors describe the product while the surveys count something else. What those breakdowns do show is that the functions with the highest reported use are office functions, where the output is a quote, an invoice, a marketing email or a booked job rather than a repair.

Trying it and running on it are different states

Adoption is not one state, and most of the disagreement across these surveys is really a disagreement about which state counts.

ServiceTitan separates the states explicitly in its 2026 State of AI in the Trades: “Two-thirds of contractors (66%) expect AI to bring moderate or major transformation to their businesses within one to three years. But adoption hasn’t caught up to that expectation yet. Only 12% have embedded AI into their operations today, and 34% are actively experimenting.” Those two shares add to 46 percent. A publisher who asked only whether the business uses AI would have collected most of that 46 percent as a yes.

Dodge Construction Network surveyed 235 general and trade contractors with CMiC, a construction management software company, in September and October 2025, and Dodge summarizes the result as “87% expect transformation, yet only 19% have adapted workflows.” The Construction Management Association of America, writing about the same study, reports that 87 percent of contractors expect a meaningful effect on the industry rather than on their own business.

Construction Dive, writing about the same survey, put it as “51% of respondents are actively evaluating several potential AI-related changes across their teams, while 40% are currently allocating a dedicated AI budget.” Expecting transformation is not use. Evaluating is real activity and budgeting is real activity, and neither one is use either.

Where publishers ask people who already use AI what they got, the answers come back positive and every one of them is self-reported, though no two publishers measure the same thing. Houzz reports that “80% of construction firms that use AI do so daily,” and that “Construction firms using AI reported saving an average of 4.7 hours per week, while 32% said the technology saves them eight hours or more.”

ServiceTitan reports early adopters “48% reporting increased productivity and 45% reporting time savings,” alongside barriers: “More than half report uncertainty around where to start, and others cite trust, cost, and system limitations as barriers.” In its commercial report, ServiceTitan states that “38% of contractors now report measurable business impact from AI, up from 17% in 2025.”

Read the benefit claims as a different class from the adoption rates. The person who answered produced that hours-saved estimate, and answered a survey run by or for a company selling into the category. The estimate is still worth something. Read it as a self-reported outcome rather than a measured one, because no publisher in this post ran a time study.

How to read the next adoption rate you see

Ask six questions of any adoption rate you meet this year and you will know much more about what it measures. None of the six needs a statistics background, and the answers are spread across the methodology note, the questionnaire and the charts, when the publisher posts them at all.

Who published it, and what do they sell? A software company reporting on its own market is not disqualified, and it does have a stake in the market it is describing. Read the rate as accurate and the publisher as interested at the same time.
Who answered, and how were they reached? Owners and executives reached through a software platform are a different population from all licensed contractors. When a publisher does not say how respondents were found, that omission is itself information.
What exactly was the question? The Census Bureau changed nothing but the wording of one phrase, and a Federal Reserve Board policy analyst reading the two series put the gap at 7.5 to 10.6 percentage points depending on the sector.
Does the summary match the question? AGC and Sage’s headline sentence describes firms using AI or planning to increase investment in it, while their published Question 15 asked whether the firm had initiated or increased AI use in the past year. When both are available, the questionnaire is the one that shows you what was counted.
What is the denominator? A percentage of AI users is not a percentage of contractors. When Houzz reports 80 percent daily use, that 80 percent covers the firms that already adopted rather than the industry.
When was it collected, and is the subsample published? A rate collected in December 2025 and published in mid-2026 covers 2025, and a single-trade rate inside a mixed-trade survey depends on a subsample whose size may not appear anywhere.

My take

I have a stake in this. We build an AI tool for technicians at Prentis, so a post about adoption rates is a post about my own category, and you should read it that way.

What the data changed for me is which rate I care about. The share of shops with AI somewhere in the building is close to meaningless for anybody building for the field, because the function breakdowns are full of office work and carry no line for a technician on a call. The rate I would rather see is how many technicians open something on a service call and still have it open at the end of the call. None of the surveys in this post publishes that. Until one does, a headline adoption rate for this industry means a shop uses AI somewhere in its business, and not that anybody used it at the equipment.

Common mistakes

Read the methodology paragraph before the headline rate, because who was asked and what they were asked separate an under-10-percent rate from an 81.5 percent rate inside the same nine months.
Read the questionnaire when the publisher posts one, because AGC and Sage’s headline sentence describes use or planned investment while their Question 15 asked whether the firm had initiated or increased AI use in the past year.
Check whether a percentage covers all contractors or only the contractors who already adopted, since the 80 percent daily use Houzz reports covers the construction firms that already use AI.
Treat hours-saved and productivity results as self-reported by the people surveyed, including the average 4.7 hours per week Houzz reports and the 48 percent of early adopters ServiceTitan reports citing increased productivity.
Ask what a tool would change about the way your technicians work, because the surveys here count office functions and never ask whether anything changed at the equipment.
Look up the subsample size before quoting a single-trade rate, since Jobber’s 81.5 percent for HVAC comes from a survey of 1,050 owners spread across eight trades.

Don’t

Do not compare a Census Bureau rate against a trade survey rate as though they measure one thing, because the Census Bureau’s survey represents employer businesses across the economy while the trade surveys reach narrower industry panels recruited by research firms like Conjointly and Thrive Analytics.
Do not read two reports from one publisher as agreement or as contradiction, since ServiceTitan reporting about 25 percent currently using AI in one report and 12 percent embedded in another is one company measuring two different states.
Do not read “percent” and “percentage points” as one unit, because a Federal Reserve Board policy analyst describes the same manufacturing shift as both 159 percent and 7.5 percentage points.
Do not buy seats for a whole crew on the strength of an industry adoption rate, because an industry rate covers what other shops told a survey and not what your technicians will open.
Do not dismiss the category on the Census Bureau construction rate alone, since that sample represents construction employer businesses of every size, and most American firms have fewer than five employees.
Do not repeat an adoption rate without the date it was collected, because several of the 2026 rates in this post were collected in late 2025.

Frequently asked questions

What percentage of HVAC contractors use AI?

There is no single answer, and the figures published in 2026 run from under 10 percent to 81.5 percent across populations that are not the same. Jobber, surveying 1,050 home service business owners in December 2025, put HVAC highest of any trade at 81.5 percent, and does not publish how many of those 1,050 owners ran HVAC businesses. The Federal Reserve Bank of Minneapolis, reading the Census Bureau’s April 2026 AI supplement, reports that less than 10 percent of businesses in construction and three other sectors reported AI use. Both organizations describe something real, and they do not describe the same population or ask the same question.

Why do AI adoption surveys disagree so much?

Publishers differ on four things. The first is who was asked, since a survey recruited by a software company reaches a narrower group than a Census Bureau sample that follows the employer business population, which a Federal Reserve Board policy analyst describes as mostly firms with fewer than five employees. The second is what “use AI” meant in the question.

The third is whether the publisher’s summary matches its own questionnaire, which is where AGC and Sage’s 61 percent sits: the headline sentence describes firms using AI or planning to increase investment in it, and Question 15 asked whether the firm had initiated or increased AI use in the past year. The fourth is who is asking, and a Federal Reserve Board policy analyst writes that senior leaders may face pressure to report AI usage as an efficiency initiative.

What are contractors actually using AI for?

Office work, in all three surveys here that break it down. Jobber reports quoting at 54 percent, invoicing at 52 percent and business writing at 51 percent among home service owners who use AI. AGC and Sage report office and administrative functions at 45 percent and estimating at 23 percent among construction firms. Houzz reports sales and marketing at 64 percent among construction firms. None of those breakdowns has a line item for diagnosing equipment during a service call.

Is AI actually saving contractors time?

The contractors using it say so, and every one of those results is self-reported. Houzz reports construction firms using AI saving an average of 4.7 hours per week, with 32 percent saving eight hours or more. ServiceTitan reports 48 percent of early adopters citing increased productivity and 45 percent citing time savings, and separately reports 38 percent of commercial specialty contractors seeing measurable business impact from AI, up from 17 percent in 2025. Read all of that as self-reported outcomes rather than measured ones, because none of these publishers ran a time study.

Did the Census Bureau really change its AI question, and does that matter?

Yes. On November 17, 2025 the Census Bureau changed its Business Trends and Outlook Survey question from whether a business used AI “in producing goods or services” to whether it used AI “in any of its business functions.” A Federal Reserve Board policy analyst reports that adoption between the legacy series and the new series jumped by 10.6 percentage points for professional services and 7.5 percentage points for manufacturing, and the Census Bureau started a new time series, citing “a level shift observed in conjunction with the new question wording.” It matters because it puts a size on how far a rate can move when the wording moves: several percentage points, across two rounds of a survey whose question was otherwise unchanged.

Which rate should a shop owner actually use?

None of them, for a buying decision. An industry adoption rate covers what other shops told a survey, and you cannot learn from it whether a tool fits the way your shop works. The useful reading of these surveys is the function breakdown rather than the headline. Quoting, invoicing and writing are where Jobber’s home service owners report the most use, and each of them produces something a shop can count, so that is the part of the operation to look at first.

Glossary

AGC: the Associated General Contractors of America, a construction trade association that publishes an annual hiring and business outlook survey with Sage.
AI: artificial intelligence. The Census Bureau’s survey question gives businesses four examples of it: “machine learning, natural language processing, virtual agents, voice recognition.”
BTOS: the Business Trends and Outlook Survey, the Census Bureau survey of about 1.2 million businesses that carries the AI questions cited here.
Census Bureau: the federal agency that runs the Business Trends and Outlook Survey and publishes its AI adoption data by industry, state and firm size.
CMiC: a construction management software company that surveyed contractors on AI with Dodge Construction Network.
Construction Management Association of America: a professional association for construction managers that published an article on the Dodge Construction Network and CMiC research.
Denominator: the group a percentage is calculated over, such as all contractors or only the contractors who already use AI.
Dispatch: assigning technicians to service calls and routing them through the day.
Dodge Construction Network: a construction data and analytics company that surveyed 235 general and trade contractors on AI with CMiC.
Houzz: a home renovation and design platform that publishes an annual State of AI in Construction and Design report.
Jobber: a field service management platform for home service businesses that published its inaugural Home Service Trends Report in 2026.
Percentage point: the arithmetic difference between two percentages, so a move from 44 percent to 61 percent is 17 percentage points and about a 39 percent increase.
Self-reported: a result given by the person being surveyed rather than measured by the publisher.
ServiceTitan: a field service management platform for the trades that publishes several annual contractor reports, three of which appear here.
Social desirability bias: the tendency of survey respondents to give the answer they believe is expected of them.
Subsample: the portion of a survey’s respondents inside one category, such as the HVAC owners inside a survey of eight trades.
Time series: a run of measurements taken the same way over time, which a change in method can break. The Census Bureau started a new AI series rather than continuing the old one.

Drafted with AI assistance and reviewed by the author.

← All posts