AI in construction: one builder for every manager, and why that is about to change
Long read, October 2026. By Andy Pritchard MCIOB, CIOB Construction Manager of the Year Gold Medal 2019. Every figure sourced inline, with full sources at the end. Last checked 11 October 2026.
AI in construction is arriving in the office, not on site. That matters, because fewer than half of the people working in UK construction are now tradespeople, and AI is likely to make the other half far more productive.
We have built an industry with roughly one builder for every manager, professional and office worker. AI and the market are about to test that ratio, and the firms that wait to find out how will be the ones it is tested on.
The numbers in brief
- 49% of the UK construction workforce is in trades and site roles, down from 65% in 2009 (CITB, 2009 and CITB, 2026).
- 2 in 5 construction professionals use AI in their daily work, up from fewer than 1 in 10 five years ago (NBS, 2025).
- -7.3%: the change in construction output per hour from 1997 to 2020, while output per hour across the UK economy rose 28.8% (ONS, via Frontier Economics).
- 4 in 10 new entrants to construction in England are training for a professional or management role, not a trade (DfE and UCAS).
- 57% of construction vacancies are hard to fill, the highest of any sector (CITB, 2026).
How has the UK construction workforce changed since 2009?
In 2009, nearly two in three people working in UK construction were tradespeople or site operatives (CITB Construction Skills Network, 2009). Today it is fewer than half (CITB Construction Workforce Outlook 2026 to 2030).
The workforce is about the same size: 2.54 million then, 2.61 million now. What changed is the mix.
- 360,000 fewer people in trades and site roles, down 22%.
- 435,000 more people in management, professional, technical and office roles, up 49%.
One in five people in our industry now works in an office-based role. That is nearly double the number of carpenters, bricklayers and roofers combined, and almost eight office-based staff for every bricklayer (CITB, 2026).
We did not stop needing people to build things. We added layers of people to manage, report, check, chase and document the people building things. The result is an industry that now employs roughly one person in a trade for every person who isn’t.
That ratio is about to be tested by artificial intelligence, because AI is arriving in one half of the industry and not the other.
A note on the numbers: CITB’s figures count everyone outside its trades and site-based group, which includes company directors. Many of those directors run small trade businesses and may well be on the tools. The job categories were also reclassified between 2009 and 2025, so the figures are rounded. The direction of travel is not in doubt.
How is AI being used in construction today?
More than two in five construction professionals now use AI in their daily work. Five years ago it was fewer than one in ten (NBS Digital Construction Report 2025).
The uses are telling. Architects, engineers and specifiers mainly use it to:
- search for technical information
- draft and review text
- analyse project data
That is the paperwork layer. Programmes, reports, RFIs, valuations, compliance packs and minutes are made of exactly this kind of work, and it is the load that has grown alongside the office headcount for 15 years.
Adoption is uneven. A global survey of more than 2,200 construction professionals found 45% of organisations use no AI at all, and only 1% have scaled it across their projects (RICS, 2025). Even so, nearly 70% of project managers and quantity surveyors believe AI will help them deliver more value.
Is AI making construction more productive?
Not measurably, and it is worth being honest about that. Most users expect significant productivity gains (NBS, 2025). The best large study so far, linking survey data to Danish employment records, found no measurable effect on earnings or hours two years after ChatGPT launched (Humlum and Vestergaard, NBER, 2025). What it did find was employers reorganising tasks around the technology.
In the authors’ words, technological change “reshapes work well before it surfaces in earnings or hours”. That is where the professions are now: the work is changing shape, and the productivity numbers have not caught up.
What I see every day
The studies lag behind what is happening on the ground. I use AI every day. I used to manage a team of 50, and my AI agents now do a large share of the processing work that team did.
The businesses I work with have an AI agent on one screen and Excel or Project on the other. That is not a pilot. It is how the work gets done.
Those businesses will expand rapidly in price-driven markets, because they will outperform you at a considerably lower price.
So don’t mistake the slow statistics for breathing space. If you’re not using AI already, you may have already lost the race.
Will AI help the construction trades?
Not yet, and not by much. I have not found any published evidence that AI is raising productivity in the trades. CITB’s workforce outlook to 2030 does not mention AI at all (CITB, 2026).
There may be gains at the edges: better logistics, tighter sequencing, less waiting for materials or information. They are plausible, but nobody has measured them yet. Even on the most optimistic view, they are nowhere near the shift under way in the office.
AI can’t hang a door or pour a slab. The technologies that change what a site produces are physical ones, such as offsite manufacture, prefabrication and better plant, and none of them moves at the speed of software.
The long-run record shows how stubborn site productivity is. Between 1997 and 2020, output per hour across the UK economy rose 28.8%. In construction it fell 7.3% (ONS, via Frontier Economics). To be fair to the industry, construction productivity has grown slightly faster than the wider economy since the 2008 crash, but from a low base and nowhere near fast enough. Part of what holds it down is work done twice: what building errors cost is a long read of its own.
So the two halves of the industry are heading for very different futures:
- The office half is about to get much more output per person.
- The site half will get roughly what it gets today.
When one side of a fixed ratio speeds up and the other does not, the ratio has to change.
Are young people training for trades or management?
Around 4 in 10 people starting a formal route into construction in England are training for a professional or management role, not a trade.
Put another way: for every three people training for the tools, two are training for the office. And England now starts almost as many chartered surveyor apprentices as bricklayers.
| Route into construction (England) | Trades and site roles | Professional, management and technical |
|---|---|---|
| Apprenticeship starts, 2024/25 (DfE) | 25,630 | 7,640 |
| University places in construction degrees, 2024 entry (UCAS) | 9,365 | |
| University places in architecture, planning and landscape, 2024 entry (UCAS) | 5,825 |
Count apprenticeships and construction degrees (building, which covers construction management, QS and building surveying, plus civil engineering), and 40% of new entrants are heading for professional roles. Include every built-environment degree and it is 47%. In the workforce today it is just over half.
We are training the next generation in almost the same proportions as the industry we have become. The pipeline is not correcting the drift. It is locking it in.
The split is also moving the wrong way:
- Bricklaying apprenticeship starts fell 39% in three years, from 3,130 to 1,910. In 2024/25 England started 1,910 bricklayer apprentices and 1,700 chartered surveyor degree apprentices (DfE).
- QS degree apprenticeships more than doubled over the same period, from 140 to 320 (DfE).
- Management apprentices are older. 30% are aged 25 or over, against 13% of trade apprentices, and 56% of trade apprentices are under 19 (DfE). Part of the management pipeline is existing staff being trained up, quite possibly people moving off the tools.
What these numbers leave out: college courses are not counted. CITB’s last study found around 104,000 people on construction college courses, but only 41% were working in construction six months after finishing (CITB, 2017). The jump in building degrees from 2022 is almost entirely students aged 25 and over (UCAS). And 2021/22 was a post-Covid peak for trade apprenticeships, so part of the fall is coming off a high. The trade and management groupings are my own classification of the apprenticeship standards.
Why is the UK short of construction workers?
If AI lets the professions produce designs, programmes, cost plans and compliance packs faster, the bottleneck moves even more firmly onto the people who build them. And that is already the scarce half.
- 57% of construction vacancies are hard to fill, the highest of any sector (CITB Industry Picture 2026).
- The industry needs around 41,000 extra workers a year to meet forecast demand (CITB, 2026).
- Apprenticeship starts would need to almost triple to close that gap (CITB, 2026).
- Even CITB’s own forecast expects less than half of workforce growth to 2030, around 45%, to be in skilled trades (CITB, 2026).
What does the skills shortage do to construction costs?
When the work is ready and the people to build it are not, the cost of those people goes up. In 2024, the industry’s main cost forecaster expected tender prices to rise about 20% and labour costs about 16% over five years. Its chief economist named skills shortages as an upside risk to wages, and to whether projects stay viable and affordable at all (Construction Management, September 2024).
That is the squeeze. More work drawn, priced and ready, competing for a trade workforce that is not growing fast enough.
Can immigration fill the gap?
For decades, part of the answer was recruiting tradespeople from abroad. Since 22 July 2025, new skilled worker visas have generally required a role at degree level. Some construction trades, including bricklayers, roofers and carpenters, can still be sponsored through a temporary shortage list, without bringing family members. That list is due to close at the end of 2026 (Construction Management, July 2025).
Whatever your view of that policy, it means overseas recruitment is a shrinking route into the trades. The workforce we need has to come from here.
Will AI replace construction jobs?
The correction will not start with redundancies. It will start with fewer people being hired at the bottom.
The first construction job AI takes won’t be on site. It will be the graduate trainee’s.
The clearest early evidence comes from the United States. Researchers at Stanford tracked payroll data covering millions of workers. Employment of 22- to 25-year-olds in the occupations most exposed to AI is now 19% below where it would have been, while experienced workers in the same jobs show no comparable gap (Stanford Digital Economy Lab, August 2026). The fall is concentrated where AI replaces tasks rather than helping with them.
The authors are careful to call this an early signal rather than proof of cause, and it is not construction data. But the pattern is easy to recognise. The tasks AI does well are the ones we give to trainees: first drafts, data entry, document checks, meeting notes, chasing information.
Now put that next to the pipeline. Around 4 in 10 of the young people choosing construction are heading for roles whose entry-level work is the most exposed of all. Trainee quantity surveyors, assistant planners and document controllers learn their trade by doing exactly the work that is being automated first.
That raises a question the industry has not answered: if AI does the trainee’s job, how does the trainee learn? My answer is that they learn on site, sooner. That is no bad thing.
What is the right ratio of builders to managers?
A ratio of one tradesperson for every manager, professional and office worker is wrong. It is wrong intuitively, and anyone who has run a site knows it is wrong in practice. My prediction is that AI, working through a free market, corrects it, and possibly very quickly.
Measure 1: the census. People working for builders and contractors, England and Wales.
| Year | People on the tools per office role | Source |
|---|---|---|
| 1991 | 3.2 to 1 | Census 1991 |
| 2001 | 2.6 to 1 | Census 2001 |
| 2011 | 2.3 to 1 | Census 2011 |
| 2021 | 1.8 to 1 | Census 2021 |
Measure 2: CITB. The whole UK industry, including architects, surveyors and engineers in consultancies.
| Year | People on the tools per office role | Source |
|---|---|---|
| 2009 | 1.8 to 1 | CITB, 2009 |
| 2019 | 1.2 to 1 | CITB, 2019 |
| 2025 | 1 to 1 | CITB, 2026 |
CITB counts consultancies, so its ratio is always lower than the census for the same year. Never compare a figure in one table with a figure in the other. Read down each one and they tell the same story: thirty years of steady drift away from the tools.
The first milestone is back to 2 to 1, last seen between the 2011 and 2021 censuses. That needs no heroic assumptions. On the census measure, it means each manager, surveyor or planner supporting about a tenth more work on the ground. Those are exactly the kind of gains AI tools are promising the office.
The second milestone is 3 to 1, where we were in 1991. That is an industry where the people building outnumber the people managing the building by three to one, and where most professionals spend their day close to the work. Getting back to 3 to 1 means two thirds more.
Where does it stop?
That is the uncomfortable question, because a free market does not stop at a sensible ratio. If software can do the work of a second planner, nobody hires the second planner. If it can then do the work of the first, the pressure carries on. Driving out cost that does not add value is what a free market is designed to do.
There are honest reasons it may not run all the way:
- Regulation. How much of the office growth is driven by rules rather than choice, such as the extra documentation now required for higher-risk buildings? AI can absorb that work, but the market cannot remove it.
- Demand. Cheaper design and planning may mean more of it gets done, not fewer people doing it.
- Accountability. Someone still has to stand on site and own the outcome. That role does not automate.
But the direction is clear. Every point on the tables above is a ratio the market will test, one hiring decision at a time.
Training
Make the managers you already have better at the craft
Going for Gold: construction project management training on your live jobs, taught by a CIOB Gold Medal winner.
What should construction firms do about AI now?
The firms that win won’t be the ones with the biggest office. They will be the ones whose managers spend their time on site, with their teams, building.
- Find out where your professional hours go. Split them into judgement (decisions, relationships, problem solving on site) and processing (reports, chasing, re-keying, formatting). The processing column is what AI takes first.
- Measure time on site, not time logged. If a site manager spends half the day at a laptop, that is the first hour to win back.
- Rethink how trainees learn. If AI does the first-draft work, put trainees on site earlier and pair the tools with time at the workface. Don’t quietly stop hiring juniors. A firm with no juniors today has no senior managers in ten years. Learning the craft on site is what Going for Gold is built around.
- Invest in trade capacity with your supply chain partners. Main contractors mostly train managers, while specialist partners carry most of the trade training (Build UK, 2024). The scarce half of the ratio needs funding and long-term work, not just a place on a framework.
- Plan for the ratio to move. Price, resource and recruit on the assumption that each professional will support more work each year, and that the trade workforce will stay tight.
Builders build. Let the machines do the paperwork.
Three free ways to start
A conversation
Tell Andy what you want to change. No pitch, and a reply within one working day.
Book a conversationA free planning workshop
For your team, online, half a day at most, on a live job of yours.
Ask about a workshopA free programme review
Send a programme. A CIOB Gold Medal winner tells you whether the dates hold.
Ask for a reviewAI in construction: quick answers
How is AI used in construction?
Mostly in the office. Professionals use it to search technical information, draft and review documents, and analyse project data. More than 2 in 5 construction professionals now use it daily (NBS, 2025).
Will AI replace construction workers?
Not on site. There is no published evidence yet of AI raising productivity in the trades. The early pressure is on entry-level office roles: in US data, employment of 22- to 25-year-olds in the most AI-exposed jobs is 19% below trend (Stanford, 2026).
What percentage of the UK construction workforce works in trades?
About 49% in 2025, or 1.28 million of 2.61 million people. In 2009 it was 65% (CITB).
Why is there a construction skills shortage in the UK?
Too few people are entering the trades. 57% of construction vacancies are hard to fill, the industry needs around 41,000 extra workers a year, and trade apprenticeship starts in England fell 20% between 2021/22 and 2024/25 (CITB, 2026 and DfE).
Is AI making construction more productive?
In the office, users expect it to, but it has not yet shown up in measured earnings or hours (NBER, 2025). On site, construction output per hour fell 7.3% between 1997 and 2020 while output per hour across the wider economy rose 28.8% (ONS).
Can construction firms still recruit tradespeople from overseas?
Only narrowly. Since July 2025, new skilled worker visas generally need a degree-level role. Some trades remain on a temporary shortage list due to end in 2026 (Construction Management, 2025).
Sources
- CITB, Construction Skills Network 2009 to 2013, UK. 2009 workforce and occupation table.
- CITB, Construction Skills Network 2019 to 2023, UK.
- CITB, Construction Workforce Outlook 2026 to 2030, UK. 2025 workforce by occupation and the forecast to 2030.
- CITB, UK Construction Industry Picture 2026.
- CITB, Destinations of construction learners in further education. 2017.
- NBS, Digital Construction Report 2025. Press release, October 2025. More than 550 respondents.
- RICS, AI in Construction 2025. September 2025. More than 2,200 respondents worldwide.
- Humlum and Vestergaard, Large Language Models, Small Labor Market Effects. NBER Working Paper 33777, 2025. Danish data, all sectors.
- ONS, Productivity in the construction industry, UK: 2021. Figures for 1997 to 2020 as analysed by Frontier Economics, Transforming the UK construction industry.
- Department for Education, Apprenticeships 2024/25. Starts by standard and age. The trade and management groupings are Constructing Culture's classification of the construction standards.
- UCAS, undergraduate end of cycle data resources 2025. Accepted applicants by subject group, 2024 entry, English providers.
- Construction Management, five-year tender price and cost forecast. 30 September 2024.
- Construction Management, Immigration reforms: what's next for construction employers. July 2025.
- Stanford Digital Economy Lab, Canaries in the Coal Mine. Updated August 2026. US payroll data.
- Build UK, Recruiting the Next Generation. February 2024.
- Census 1991, table L76, occupation by industry, England and Wales. 10% sample. Via Nomis.
- Census 2001, table ST039, occupation by industry. Via Nomis.
- Census 2011, table DC6604EW, occupation by industry. Via Nomis.
- Census 2021, industry by occupation. ONS custom dataset.
- Ratios from the census and CITB tables are Constructing Culture calculations. On the census measure, “on the tools” means skilled trades, process, plant and machine operatives, and elementary occupations; everyone else counts as a manager, professional or office worker.
Andy Pritchard MCIOB
CIOB Gold Medal, Construction Manager of the Year 2019. Founder of Constructing Culture and author of Going for Gold. About Andy →