It’s Sunday evening and your estimator is measuring a four-storey office block off PDF drawings. The tender goes in on Tuesday. The last job anything like it was priced by someone who left in the spring, and his spreadsheet has three tabs nobody can explain.
That is where AI in demolition starts to earn its place. In plain terms, AI in demolition means software that reads the drawings, documents and project history a contractor already holds, and does the slow joining-up work a person would otherwise do by hand: measuring, cross-checking, drafting the first version. It doesn’t survey a building, sign off a RAMS or decide what goes in a tender. People still do that. What changes is how much of their week goes on typing and measuring, and how much goes on judgement.
Why has demolition been left out of the AI conversation?
Because most of the people building the software don’t understand it. Surtori’s founder, Torquil, puts it bluntly: demolition gets treated as construction’s annoying little cousin. Founders look at it and see a small market with messy data. They don’t see the structural survey that changes the whole sequence, the asbestos found behind a riser after the price has been agreed, or the waste classification that decides whether a load goes for recycling or to a hazardous site at several times the cost.
Yet the margin problems are the same ones construction has, and often sharper. Jobs are short and prices get fixed early. The information that would protect the margin sits in several places at once: the estimating file on one laptop, plant on the hire desk’s whiteboard, the RAMS for the last similar job in someone’s inbox, the programme in a separate tool. None of it talks to anything else. Most of the value AI brings to a demolition firm comes from joining that up, and very little of it comes from anything clever.
Where should a contractor start with AI in demolition?
With the tender. Estimating is the one place we can point to a measured result rather than a promise, and it’s also where the weekend hours go.
We built an AI-assisted estimating capability for a UK construction and fit-out contractor turning over around £30 million. Estimating time per bid fell by 60–70%, and cost accuracy improved, with the gap between estimated and delivered cost narrowing. The mechanism is unglamorous. The system reads the drawings and the firm’s own priced history and produces a first pass of quantities and rates. An experienced estimator checks it, changes what’s wrong and puts their name on the number. That last step never moves.
For a demolition firm the inputs are different: floor areas and structural types, soft strip quantities, arisings by waste stream, plant days, the access constraints of a tight urban site. The approach is the same. Our walkthrough of AI demolition estimating goes through how a tender gets priced this way.
Starting here has a second benefit that’s easy to miss. Clean, connected estimating data is the groundwork for everything below, because the same quantities feed the RAMS, the waste forecast and the plant plan.
Can AI write a RAMS or a Section 61 application?
It can draft one. That’s a narrower claim than it sounds, and a useful one.
Most RAMS problems aren’t problems of knowledge. They’re copy-and-paste problems: last month’s document with the old site address still in the header, a sequence written for a different building, a hazard that doesn’t exist on this job while the one that does is missing. A system working from your own approved library of activities, with the site specifics fed in, produces a first draft that starts from the right job. Your competent person then reads it properly, which is what CDM 2015 expects of them anyway. Our guide to the risk assessment method statement covers the document itself, and the POWRA guide covers the check done at the workface on the morning.
Section 61 shows where drafting help pays off, because the paperwork is heavier than the law. Section 61 of the Control of Pollution Act 1974 asks for particulars of only two things: the works and the method of carrying them out, and the steps you’ll take to minimise noise. Councils usually want a good deal more on top, typically BS 5228 predictions, a plant list and working hours, because they have to be satisfied the information is sufficient. Most of that already sits in your programme and plant schedule. Assembling it is where AI helps. Choosing the method stays with you. We cover what goes into a Section 61 application in detail, along with how the drafting can be automated.
One detail catches people out. The Act puts no duty on the applicant to tell the council when the plant changes, and a consent has no default expiry. So a consent can quietly stop describing the job you’re running now. A system that holds the consent next to the live plant list can flag the mismatch before an environmental health officer does.
What does AI do for plant and programme?
Picture two crews arriving on the same floor on the same morning. The licensed asbestos team is still finishing an enclosure, and the steel cutters were booked a week ago off an older version of the programme. Nobody made a bad decision. The information just didn’t travel. Joining the programme to subcontractor bookings and the site diary is enough to catch that clash days ahead, instead of at seven in the morning in the site car park.
Plant is the same story with money attached. The excavator that came off the job three weeks ago and is still on the hire invoice turns up on more monthly reports than anyone likes to admit. Connecting telematics and hire records to the programme shows which kit is earning and which is parked. Our piece on plant management software sets out what that system should track.
Torquil compares running a contractor without this view to running a Formula 1 team with no telemetry. The drivers are good and the car is quick, but nobody in the garage knows what’s happening until the race is over. Live dashboards across several jobs aren’t a technical challenge any more. Getting the data into one place is, which is why this comes after estimating.
Where do waste tracking and urban mining come in?
Tier-1 clients increasingly set recycling and recovery targets and ask for evidence behind the figure on the tender return. That evidence starts before the first machine arrives, with a pre-demolition audit forecasting what the building contains by material stream. AI helps turn that forecast into a live record: quantities taken from drawings or a model, tagged by stream, and tracked through to the tickets coming back from the recycler.
Regulation is pushing the same way. Digital waste tracking becomes mandatory for permitted waste receiving sites in England and Wales from October 2026, with Scotland and Northern Ireland following in January 2027 and carriers and brokers later that year. Firms whose waste records are already digital will find the switch far less painful. Urban mining, treating the building as a stock of materials with resale value, is the longer view, though reuse markets for most non-metal components are still thin. Treat it as cost control and evidence first and resale second. Hazardous material is where the money moves fastest, and our breakdown of asbestos and demolition costs shows how survey findings feed the price.
How far along is each use?
Only estimating has a measured client result behind it. The others are capabilities we can build and demonstrate, but can’t yet put a client figure against.
| Use | What it does for you | What’s behind it today |
|---|---|---|
| Estimating and takeoff | First-pass quantities and rates from drawings and your own priced history | Measured client result (see above) |
| RAMS and POWRA | First drafts from your approved activity library, with site specifics fed in | Working capability, no client figure yet |
| Section 61 and compliance packs | Assembles the application from programme and plant data you already hold | Working capability, no client figure yet |
| Plant utilisation | Shows idle and double-hired kit across sites | Capability; depends on telematics and hire records being available |
| Programme and clash checks | Flags trades booked into the same space on the same day | Capability; depends on how the programme is kept |
| Waste tracking and urban mining | Forecasts arisings by stream and tracks them through to recycler tickets | Capability; digital waste tracking adds pull from October 2026 |
Questions contractors ask us
Will AI replace demolition estimators?
No. It removes the measuring, not the estimator. Someone still has to know that a 1960s frame will throw up something the drawings don’t show, and price for it.
Do we have to replace Sage, Procore or the systems we already run?
No. The work sits on top of what you have and joins it up. It’s built inside your own Microsoft or Google cloud account, so the data stays in your tenancy and we neither host it nor own it.
Is it safe to put tender and project data into AI tools?
That depends on where the data goes. Pasting a priced bill of quantities into a public chatbot is a very different thing from a model running inside your own cloud, under your own access controls. We build the second kind, and a person signs off every output that carries liability.
How long before we see anything?
About ten days for the first piece. For an average-sized firm, that’s how long the initial needs assessment takes, and it ends with a map of where your data lives and a roadmap ranked by return. Estimating is usually near the top. After that the pace and the budget are yours to set.
Is AI in demolition different from AI in construction?
Mostly in the inputs. Asbestos, structural surveys, soft strip and waste classification change the data a job depends on, and tools built for new-build tend to miss them. Our work across the wider sector is on the construction AI advisory page.
One tender at a time
Torquil ends most conversations on this subject the same way: you won’t lose your job to AI, you’ll lose it to someone who uses it. For a demolition contractor, using it well looks less like a transformation programme and more like pricing next month’s tender without losing the weekend to it, with the same person accountable for the number.
Our demolition contractors page explains how we work with firms like yours, and Fusion for demolition contractors shows the software that does the work. If you’d like to find out where your first tender would be, book a discovery call.