AI in construction

Why AI in construction starts in the office, not on the site

Construction companies are sold AI from the site side: drones, progress recognition from photos, "AI plus BIM". Those projects can be valuable, but they share a precondition we check before starting — orderly, systematically collected site data, which most companies simply do not have in that form. The project then starts with setting up data collection, and the first result shows after a year, not a quarter.

Meanwhile, in the same company, a stream of work flows daily that AI can take over today: tender documentation running to hundreds of pages, read under deadline pressure; purchase invoices assigned to sites and stages by hand; subcontractor invoices checked against contracts and work progress records one by one; project correspondence in which questions and agreements get lost between inboxes. These are document- and text-based processes — the data exists, the rules can be written down, and a mistake is detectable and reversible.

That is why a good first AI project in a construction company looks unspectacular: it does not touch the site, it touches the flow of information around the sites. In return, the first results appear within weeks, which helps fund the next steps.

What we actually automate in construction companies

• Tender documentation analysis. Extracting scopes, deadlines, contractual penalties and payment terms from specifications and bills of quantities — as an organised summary that the estimator and the bid decision start from. It shortens the stretch where today the "do we bid" decision is made after a skim, because there is no time to read the whole thing.
• Purchase-invoice processing. Reading invoices arriving by email and on paper, assigning them to the right site and stage based on orders and contracts, and preparing the posting for approval. Ambiguous cases go onto an exception list for a person — not straight into the books.
• Checking subcontractor invoices. Comparing subcontractor invoices against contracts, amendments and work progress records, with a list of discrepancies to resolve — instead of comparing three documents by hand at every payment. A missed discrepancy can mean an overpayment, so the list of differences is reviewed before the payment is approved.
• Handover records and as-built documentation. Tracking the completeness of the documents required for hand-overs and payments — certificates, declarations, subcontractor statements — with a list of gaps and prepared reminders, before a gap stops an acceptance.
• Project correspondence. Organising questions, agreements and changes from emails into one register assigned to the site, with a "who agreed what, when" history — plus an assistant that answers recurring questions from the project documentation, citing its source.

The common denominator: a person approves the result, the system prepares it. Which processes qualify for this kind of automation at all — the five conditions a good candidate must meet — is described on the AI automation page; on construction projects those conditions are checked exactly the same way.

What we do not promise

• An estimate without an estimator. AI prepares the material — scopes from the documentation, comparisons with earlier quotes — but the valuation, its assumptions and its risks are the responsibility of a person who knows the technology and the local rates. An offer promising an "automatic estimate" is promising decisions with nobody accountable for them.
• Site analytics "on the side". Progress recognition from photos or safety monitoring first require systematic image collection and a separate project with its own budget — also because site footage is employee data, with its own legal constraints. It can be done well, but not as an add-on to document automation.
• A workflow without oversight from day one. Everything that affects payments and obligations — postings, reconciliations, replies to subcontractors — first goes through a stage in which a person approves every result. Loosening the oversight is a decision made on pilot data, not on assumption.

How we work

As in every project: we pick one process and measure one indicator before we start, for example hours from invoice arrival to a ready posting. Then we run the smallest possible pilot. Before switching over, the system runs alongside the current way of working for a while. At first the system prepares and a person approves every result. The full course — four phases and what you get at the end of each — is described by the delivery methodology, and the full scope of the service, from the process audit to maintenance, by the AI implementation page.

A separate topic that comes back in every construction conversation is data: cost estimates and rates, subcontractor agreements, personal data in licence documents and statements. Where data goes when we integrate an AI model, and what to ask any provider, is set out on the security and GDPR page.

What it costs

Three things drive the cost: the number of systems to connect, the state of the data, and how reliable the results must be, that is, how rarely they may be wrong. The number of sites matters less — what counts is whether documents arrive through one channel or into each site manager’s inbox separately. All the components, maintenance included, are broken down in our article on what an AI implementation costs.

Frequently asked questions

Where should a construction company start with AI?

In the site office, not on the site. Analysing tender documentation, processing purchase invoices, checking subcontractor invoices and payments, and keeping project correspondence in order are document-based processes — the data already exists in mailboxes, PDFs and the accounting system, a mistake is reversible, and results show in weeks. Projects built on site data — photo analysis, progress tracking from cameras, drones — first require setting up data collection, so they cost more and take longer. Starting with them is the most common way construction companies spend their budget before seeing a first result.

Will AI prepare our cost estimates?

No — and an offer that promises this should raise a flag. An estimate is the estimator’s responsibility: it takes knowledge of the technology, local rates and risks that are not in the documentation. What AI does is shorten the road to the estimate: it extracts scopes and line items from the tender documentation, compares an enquiry against earlier quotes for similar work, and prepares the material the estimator starts from — instead of starting from a hundred-page PDF. The decision and the signature stay with a person.

Will this work with our estimating and accounting software?

The audit at the start of a project settles that. We check whether the API (the interface other software uses to connect to the system) or file import gives access to the data and actions the chosen process needs. On that basis we confirm whether the integration is feasible, what it covers and what it costs — and whether an existing system feature would be enough. In construction companies the same work is often spread between the estimating package, the accounting system and per-site spreadsheets — and it is precisely the data copied manually between those systems that we automate first.

How much does an AI implementation cost in a construction company?

The cost is driven by the number of systems to connect — email, accounting, the estimating package, per-site spreadsheets — the state of the data, and how reliable the results must be, that is, how rarely they may be wrong. The number of parallel sites matters less than whether documents arrive through one channel or into each site manager’s inbox separately. Maintenance is a separate line: usage-dependent fees and periodic tests after model version changes. We broke the cost components down in our article on what an AI implementation costs.

What about the data in tenders and subcontractor agreements?

Tender documentation, cost estimates and subcontractor agreements are the information your margin depends on — and they also contain personal data: contacts, licence details, sometimes subcontractors’ employee data. Before the integration, not after it, you need to settle which data leaves the company at all, where it goes and on what basis — from the choice of provider and processing region to limiting the data to the minimum a given step needs. The rules, and the questions to put to any provider, are set out on our security and GDPR page.