Construction runs on paperwork that arrives in every format a supplier, a site or a designer can produce, and on processes that cross two teams before anyone can price them. It is the sector where AI has the most to remove and the least software that fits.
Agency AI is an AI consultancy in Glasgow that builds AI for construction companies across the UK. Inside a construction software platform, our AI reads and matches procurement paperwork in under five seconds a document. For a groundworks contractor, we built an operating system for variations, material requests and planning. For another contractor, bid writing time fell by 90%.
Every other industry gets software built for it. Construction gets software built for offices, then spends years bending its processes to fit, or running the real work in email alongside the tool it pays for.
Three things make it different. The paperwork arrives in whatever format the sender had to hand, including photographs of printouts. The processes cross teams and come back on themselves, because a variation needs the person who raised it before it can be priced. And the value of the job changes the route it should take, so a rule that works at four thousand pounds is wrong at forty.
That is exactly the ground AI covers well, and it is why construction is where most of our work sits.
Delivery notes, invoices and credit notes are read and matched against your own records as they arrive, then totalled, rather than piling up until someone has an afternoon.
Variations, material requests and approvals run as processes that know which team owns each step, what the value threshold is, and when something has gone quiet.
Drawings, schedules, specifications and contracts stop being a folder nobody searches and become something you can ask a question of.
Different businesses, same six conversations.
Ordered, delivered, invoiced and credited all live in different places, so the commercial position on a project is assembled by hand and is out of date by the time it is finished.
The tool records the decision but does not run the process, so the real work happens in inboxes and nobody can say what is open or who it is sitting with.
The answers exist in submissions already written and won, and they get rewritten anyway because finding them takes longer than retyping them.
The same measures, rates and checks are repeated on every job, by the people who are most expensive to have doing repetitive work.
Applications arrive in bursts, get re-keyed by hand, and the candidate has three other offers before anyone replies.
Drawings, specifications and contracts hold the answer, and finding it means somebody senior reading for an hour.
Hundreds of delivery notes, invoices and credit notes uploaded at once, each read and matched against company records in under five seconds, then totalled against the budget. Read the case study →
Variations, material requests and approvals running as real processes: owned by teams, routed by value, sent back when they need to be, and flagged when they stall. Read the case study →
CVs read by AI as they arrive, filed against the candidate, scored, then email and SMS to the candidate and a booking link they use themselves. Read the case study →
Measures, rates and catalogue pricing brought together so a quote is assembled from the business's own numbers rather than rebuilt each time. Read the case study →
First drafts written from the library of submissions already won, in a consistent voice, with the knowledge from past bids kept rather than lost. Read the case study →
Structured data pulled out of drawings and schedules, so what is on the sheet becomes something the rest of the system can use. Read the case study →
Two versions of a specification compared, and contracts read for the risk that matters, in minutes rather than an afternoon. Read the case study →
Ask a question and get the answer with the document it came from, instead of opening folders.
We build the AI inside APSIS Business Components, a construction software platform, where the work has run for over a year and new capability ships most weeks. And we built M Squared, a commercial landscape and groundworks contractor named in the Financial Times' 1,000 fastest growing companies in Europe, their own operating system.
Both are partnerships rather than deliveries, which is the only way this work holds up: the processes change, the business grows, and the system has to move with it.
The work that repeats and the work that waits. Reading procurement paperwork and totalling it against the budget, running variations and material requests as processes that cross teams, drafting bids from submissions already won, pulling data out of drawings and schedules, and answering questions from project documents.
Yes. We build alongside the systems already in place rather than replacing them, and everything we build feeds back into them. Where a process genuinely does not fit any product on the market, we build that part bespoke and the software belongs to the business.
Our work is built for established mid-market businesses, the kind where manual processes have grown with the company and now carry real cost. If a handful of processes are consuming whole roles, that is the profile.
Tell us where your team loses the most time. We will tell you honestly whether AI pays there, what it takes to build, and what we have already delivered for businesses like yours.
We use cookies to measure how the site is used and, if you allow it, for advertising. Nothing is set until you choose. Privacy policy.