AI automation · Construction Construction AI and automation for contractors: from site diaries to contract notices

Typical workflows in a construction company

These are typical processes, not client projects. In each, the AI prepares and a site manager or commercial manager decides.

1. Site diaries and progress reports

Site diaries are written at the end of a long day, and progress reports are assembled from them at the end of the month.

• Input: voice notes, photos with location, weather, labour and plant records and the programme.
• AI prepares: the daily record with work done, labour, delays and instructions received, and the monthly progress report with photos placed against the right activity, flagging gaps such as a day without an entry.
• Site manager checks: facts and wording, because the record may be used later as evidence.
• Output: daily records and a report draft in the company’s layout.

2. Quantities and subcontractor quotes

Quotes arrive from subcontractors in different formats, with different scopes and different exclusions.

• Input: drawings and specification, the bill of quantities, subcontractor quotes.
• AI prepares: reads each quote into a common structure, sets scope, rates, exclusions and qualifications side by side and highlights items that are missing or priced differently. AI‑assisted measurement tools can propose quantities from drawings, and the AI collects them in the bill.
• Quantity surveyor checks: the comparison, every measured quantity that carries a payment and the recommendation.
• Output: a leveled comparison and a quantities draft with references to the drawings.

3. Instructions, variations and notices

Instructions arrive by email, in meetings and on site. Delays and extra work have to be notified, often within a fixed period.

• Input: emails, minutes, instructions and the programme, and the contract’s notice provisions entered once by the commercial team.
• AI prepares: a log that links each event to its source, a reminder before each notice period ends and a draft notice that quotes the facts and the contract clause.
• Commercial manager checks: whether a notice is needed, its content and its date, and signs.
• Output: an event log, a deadline list and notice drafts.

4. Subcontractor and safety documents

Before a subcontractor starts, risk assessments, method statements, insurance and certificates must be present and in date.

• Input: the documents the subcontractor sends and the company’s requirements.
• AI prepares: checks each document against the requirements, lists what is missing or expired and drafts the request to the subcontractor. It tracks expiry dates and reminds you before they pass.
• Competent person checks: the content of method statements and risk assessments and approves the start.
• Output: a checked document file per subcontractor and a list of expiry dates.

What stays with site managers, and what we do not do

• Safety and engineering: decisions about safe methods, temporary works and structural matters stay with competent people.
• Measurement and payment: quantities, valuations and payment notices are checked and issued by the quantity surveyor or commercial manager.
• Notices and claims: the AI reminds and drafts. A person decides whether to serve a notice and signs it.
• Site control systems: we automate documents and communication, not machine control or safety functions.
• Decisions about people: we do not build systems that rate, monitor or select individual workers. These are high-risk under Annex III of the AI Act, see risk categories.

Rules that matter

The table lists the main sources as an orientation. Scotland and Northern Ireland have their own variants, and every contract has its own terms.

On top of this, the GDPR applies to personal data on site, for example workers’ records and photos with people on them. You need a data processing agreement with every provider (Art. 28), servers in the EU or UK and a clear rule on how long photos are kept. Art. 4 of the AI Act asks for sufficient AI literacy among staff, and Art. 50 requires AI chatbots to identify themselves. This overview is general information, not legal advice. Your contract documents, your solicitor and your trade body can confirm what applies.

Systems you probably use

Contractors typically work with project and document platforms such as Procore or Autodesk Construction Cloud, planning tools such as Asta Powerproject, Primavera P6 or Microsoft Project, measurement and mark-up tools such as Bluebeam, and commercial software for valuations and accounts. In German-speaking countries, tendering and cost software such as ORCA AVA, RIB iTWO or NEVARIS is common. We connect through exports, the interfaces these systems offer, email and shared folders. The names are for orientation only and do not imply any partnership.

How AI assistants find contractors

Clients and main contractors ask ChatGPT, Gemini or Google’s AI Overviews for “fit-out contractor for offices in Manchester” or “groundworks contractor in Cork”. The answer draws on the company’s own pages, accreditation and trade body listings, directories, project pages and reviews.

• A page per work type: groundworks, fit-out, refurbishment, new build, with sectors, typical project size and how a project runs.
• Project pages with substance: location, scope, constraints and outcome. Client names appear only with the client’s consent.
• Accreditation entries that match: company name, address and numbers as listed by schemes such as Constructionline, CHAS or SMAS in the UK.
• The same details everywhere: website, Google Business Profile and directories.

Is it worth automating? Eight questions

1. Are site diaries written at the end of the week from memory?
2. Does the monthly report take days of collecting photos and notes?
3. Do you compare subcontractor quotes in a spreadsheet that someone builds from scratch each time?
4. Are notices and time limits tracked in someone’s head?
5. Do you chase subcontractors for insurance and method statements by email?
6. Does the same information sit in email, the platform and spreadsheets?
7. Is there a written rule on what may go into AI tools and who checks the result?
8. Can you name one task where a draft that needs checking would already save time?

Three or more yes answers usually point to a workflow worth testing. Question 7 is the prerequisite.