I've been using software on construction jobs since the late 90s, when planning software was still a novelty on site. In 2001, I built a piece of software called Powerboard. It managed plant: which machine was where, on which job, booked to whom. I knew exactly what it needed to do, because I had spent years running it with a manual T-card system on my wall that tracked the locations of all the plant. Making that physical thing, the one I stared at for hours a week, into something on a screen was the dream.
In those days it was Visual Basic 6 and an SQL database, with everything held locally on the office network. Development was slow and expensive: the developers, the months of build, the money to keep going until the industry caught up. We built it, people used it, and some groundworkers still do. But we ran out of runway before the market was ready, and I learned the lesson every tradesman with a good idea learns sooner or later. Knowing the problem was never the barrier. Building the thing was.
That barrier has all but gone.
The cost of building has collapsed
Writing software used to be the slow, expensive part. Now a large share of it is written by AI. In the startup world that changed almost overnight: Y Combinator, the best-known start-up accelerator, said in early 2025 that a quarter of its latest batch had codebases that were around 95% AI-generated, and these were not non-technical founders. The shape of who starts companies is changing with it. Carta, which tracks tens of thousands of US start-ups, found the share of new companies started by one person rose from under a quarter in 2019 to more than a third by mid-2025.
The same thing is happening inside firms. McKinsey's 2026 survey found roughly a third of organisations had skipped buying a piece of software at least once because they could build it themselves with AI coding tools.
Put simply: what took me a team and a bank loan in 2001, someone with the same site knowledge can now get to a working version largely on their own.
Who wins
The estimator who has rebuilt the same spreadsheet on every tender for twenty years. The planner who knows exactly why the programme slips at first fix. The site manager who has watched three apps fail to do what a whiteboard does. For most of the history of software, those people could describe the tool they needed perfectly and still never see it built, because the people who could build it had never stood on a slab at seven in the morning.
Deep industry knowledge was always the scarce part of good software. Now it is very nearly the only scarce part. That is why I'm optimistic, and after more than 25 years of using technology on construction jobs, I don't say that lightly. Construction has more of that knowledge than almost any industry, and most of it has never been written down, let alone turned into a product.
What the landscape looks like
Expect a long tail of small, sharp tools. One job, one trade, sometimes one type of contract. Priced like a subscription to a trade magazine, not like an enterprise system. Built by someone who has done the work and is still close enough to it to answer the phone when it goes wrong.
Some will be brilliant. Some will be a spreadsheet with a login. Most of them will arrive at once.
What didn't get cheaper
This is the part the excitement leaves out, and it matters.
Building the first version got cheap. Running it for years did not. A careful study by the research group METR in 2025 found that experienced developers working on large, mature codebases were actually around 19% slower with AI tools, while believing they were faster. A new tool is easy to start. Keeping one secure, supported and working as it grows is still hard work.
I know this first hand. Yes, it's easier. But even a small product still takes a database (Supabase), a web framework (Next.js), Python, somewhere to host it (Vercel, Railway), somewhere to keep the code (GitHub), and a whole bunch of secrets that need to stay secret. Then you have to plumb it all together, and keep it working and secured, day and night.
And the things that decide whether a contractor can rely on a tool are mostly not code:
- Trust: will it still be here when the job finishes, and who has my data if it isn't?
- Support: who answers when it stops working on a Friday afternoon with a pour booked for Monday?
- Fit: does it talk to the systems the firm already runs, or is it one more login and one more place to type the same thing twice?
- Selling: construction buys carefully, and a better tool still has to get past the person who signs the order.
- Staying in business: plenty of these tools will disappear. It's why the Redstick record now notes, with a date, when a tool is withdrawn or its maker closes.
What it means if you run jobs
More choice, and a lot more noise. The answer isn't to wait it out; some of these tools will save real money. It's to buy the way you would buy plant or a subcontract:
1. Start from the job, not the product. What exactly do you need done, and what do you do now? 2. Ask the maker the hard questions. Who runs it on site? What has to be in place before it arrives? What does it replace, and what does it add? 3. Check they'll still be there. Who are they, how long have they been trading, who else is running it, and what happens to your data if they go?
And if you're tempted to build your own, and some of you will be, keep that plumbing in mind. Writing the tool is only the start. The question only you can answer is whether building and running it is the best use of your time, or whether a third-party app that costs about the same each month as a meal for two is the better option.
The upside
For years the tools we used on site were mostly designed by people who had never worked on one. That is about to change. The next wave of construction software will be written by estimators, planners, engineers and site managers: people who have stood in the mud and know what the work actually needs.
That's a good thing for the industry. It's also why somebody needs to keep a straight, dated record of what's out there, so the good ones get found.
John
Sources (read 8 Nov 2026)
- Y Combinator W25, a quarter of the batch with ~95% AI-generated codebases (Jared Friedman, YC): TechCrunch, 6 Mar 2025, techcrunch.com/2025/03/06/a-quarter-of-startups-in-ycs-current-cohort-have-codebases-that-are-almost-entirely-ai-generated
- Solo founders 23.7% (2019) to 36.3% (H1 2025) of new US start-ups: Carta, Solo Founders Report 2025, 9 Dec 2025, carta.com/data/solo-founders-report
- ~32% of organisations skipped a software purchase to build with agentic coding: McKinsey, The state of AI in 2026: On the road to ROI (25 Aug 2026; 1,719 respondents in 97 countries, fielded 4 May to 8 Jun 2026), read at source 9 Nov 2026: "Nearly a third of respondents (32 percent) report that their organizations have decided against buying one or more software products or features because they could be built internally with agentic coding tools." www.mckinsey.com/the-state-of-ai
- Experienced developers 19% slower with early-2025 AI tools on mature repositories: METR, Becker et al., Jul 2025, arxiv.org/abs/2507.09089