In most building and trade businesses the paperwork runs a step behind the site. Drawings and scopes arrive by email, the estimator prices them after hours, supervisors send job sheets and photos from their phones, and someone in the office turns it all into claims, certificates and invoices. Subcontractor insurances sit in a folder until one lapses. We automate the predictable steps as fixed, tested software that applies your rates, deadlines and checklists the same way every time. AI is used only where something has to be read, such as a scope, a variation email or a photo of a licence card, and its output is checked before anything leaves the system.
These are patterns we see in construction and trades businesses, not client case studies. Your process gets its own map in the assessment.
Getting quotes out before the builder moves on
A request for quote lands as an email with a PDF drawing set, a scope from a builder, or a voicemail the office types up. The estimator reads it, measures off the plans, pulls rates from an old spreadsheet and builds the quote. AI reads the request and drafts the quote lines, listing the inclusions, the exclusions and anything it could not find in the scope. The pricing underneath comes from your rate library, labour allowances, margin rules and quote template, applied the same way every time, and a follow-up schedule chases quotes that go quiet. Where the plans are unclear, take-off stays with the estimator, who checks every line and approves every quote before it is sent.
Progress claims lodged on time, every month
Missing the claim date, or leaving out a variation, means the cash waits another month. A progress claim is built from the contract schedule, the progress on each line and the approved variations, then sent before the cut-off the contract sets. The claim build is automated, carrying forward previous claims and retention and diarising the dates set by your contracts and the security-of-payment rules. Variation instructions arrive as emails, site diary notes or a text from the site manager, so AI reads that correspondence and lists each possible variation with its source. The project manager decides what is claimed and approves every claim before it is issued. Interpreting the contract stays with you and your advisers.
Tenders, insurers and the inbox
Head contractors, councils and developers each send their own tender and prequalification forms. Each asks for the same company details, WHS system, insurances, project history and key personnel, in a different order and a different template. Somebody copies last time’s answers, updates what changed and hopes nothing is stale. We build a library of approved answers and current documents, and a compliance matrix that tracks every returnable schedule against the tender’s requirements, and assembly and formatting are automated. AI drafts answers to new questions from the approved library only, and flags any question it has no approved answer for. The methodology, the price and the decision to bid stay with you, and whoever you nominate signs off every submission.
Insurance repair builders and restoration trades work to the insurer’s timeframes. A claim arrives as a scope of works from the assessor, and what follows is a run of site attendance reports, photos, quotes, portal updates and homeowner phone calls, with the insurer watching how long each step takes. The claim checklist, the status updates to the insurer’s portal or inbox and the homeowner’s notice at each stage all run automatically. AI reads the loss description and the assessor’s scope, and drafts the first site report from the supervisor’s notes and photos for the supervisor to correct. Scope disputes, make-safe calls and anything the insurer queries go to your claims coordinator.
The office@ and accounts@ inboxes of a building business collect everything. Quote requests, supplier bills, variations, council notices, complaints, certificates and spam land together, and the urgent ones wait behind the routine ones. AI reads each message and attachment, classifies it and pulls out the job number, the sender and any date it mentions. Fixed routing rules then send each item to the right person or into the right process, so quote requests reach the estimator and bills reach accounts. When the AI is not confident, the message goes to a review queue in the office instead of guessing. This inbox is often the front door for the other processes on this page.
Closing out jobs and getting paid
In a trade business the invoice depends on what comes back from site. Timesheets, materials drawn from the van, extra work agreed on the day and the technician’s notes all have to be gathered before anyone can bill. When they arrive late, the invoice waits, and the small extras are never charged at all. Each invoice is built from the source records, taking labour from timesheets, materials from supplier dockets and stock issued, and variations from approved changes. Each is checked for completeness against the job before it reaches accounts. AI has a small part, reading free-text job notes for extra work the technician mentioned but did not record. Anything uncertain goes to the office manager first.
A finished electrical or plumbing job leaves a trail of paper. The job sheet, photos, test results, compliance certificate and customer sign-off all have to be kept. Field-service software already handles much of this well, and where one app holds the whole job, we would say so. The gap is where the paperwork spans several places, with photos in a phone gallery, certificates in a regulator’s portal and the invoice in the accounting system. Fixed software checks each job for completeness, assembles the certificate from the recorded test results and hands the finished job to billing. AI reads the technician’s notes and photos to spot what is missing. The licensed electrician or plumber still signs every certificate.
Where your contracts or state rules require a project trust or retention trust account, every deposit and payment has to be recorded against the right project and subcontractor, and the account reconciled on the schedule those rules set. Mistakes here carry serious consequences, so the job usually falls to the owner or financial controller late in the month. No AI is involved. Bank transactions are matched to claims and subcontractor payments, a ledger is kept per beneficiary, and the reconciliation is prepared with anything that does not balance listed. We handle the records and the routine steps. Your financial controller reviews and signs off every reconciliation, and the business keeps responsibility for meeting its trust obligations.
Ordering from and paying suppliers
Late trusses, windows or switchboards can hold up the whole trade sequence behind them. The purchasing officer or site manager raises the order, emails it, rings to confirm the date, then rings again when the date moves. Supplier replies come in every format imaginable, so AI reads them for delivery dates, substitutions and short supply. Everything around that follows set rules. Purchase orders are raised from the approved job budget and the programme, sent and chased until each supplier confirms, and the confirmed date is written back against the job. When a changed date clashes with the programme, the site manager decides what moves.
Bills for concrete, steel, timber and plant hire arrive by email, while the delivery docket that proves what arrived was signed on site and often never reaches the office. Each bill has to be matched to a purchase order and a job, coded, and approved by the project manager who ordered the goods. Receipt capture tools already read most invoices well, and we use them where they work. What stays manual is the exceptions. Matching bill, order and docket within the price and quantity tolerances you set is rule work, as are chasing approvals and reconciling supplier statements at month end. AI reads only the invoices and dockets the capture tool cannot. It pays for itself when supplier volumes are high.
Crews, subcontractors, plant and safety
Before a subcontractor sets foot on site you need their public liability and workers compensation certificates, their licence and a SWMS for the work, all of it current. Most builders keep this in a spreadsheet and a folder of PDFs, and find the lapsed policy after something goes wrong. Packaged prequalification platforms exist, and if your subcontractors use one, we connect to it. Otherwise we build the register, the expiry alerts and the chasing. AI reads each certificate of currency for the insurer, policy dates and cover limits, and a reading that fails the checks is held for review before the register changes. A lapsed policy is flagged to the site manager before the subcontractor’s next booked day.
Your crews carry white cards, high-risk work licences, EWP and forklift tickets, first aid certificates and site inductions, each with its own expiry. A head contractor will turn a worker away at the gate for a lapsed one. The register, when it exists, lives with whoever writes the roster. We keep it as a live register, with alerts to the worker and supervisor ahead of each expiry and a block that stops anyone being allocated to a job that needs a ticket they do not hold. Renewed cards are photographed and sent in. AI reads the card for the licence class and dates, and the supervisor confirms the reading against the photo before the record changes.
On site, a near miss is phoned through to the supervisor, written on a form in the ute, or never written down. By the time it reaches the WHS coordinator the details are thin and the follow-up actions have no owner. Fixed software gives the crew one short form on their phone, routes each report to the right supervisor, sets deadlines for each corrective action and sends anything serious to the WHS coordinator straight away. Psychosocial hazards, such as overtime fatigue or a bullying complaint, go into the same register with restricted access. AI summarises the free-text report and photos for the register. Every investigation, and every call on whether the regulator must be notified, stays with your people.
Between the utes, a tipper, an excavator and a few scissor lifts, a trade business carries a long list of registration dates, service intervals and plant inspections. That list tends to live in the fleet manager’s head or on a whiteboard in the yard, and a lapsed rego or a missed inspection shows up at the worst time. Most of this needs no AI. The register reads hours and odometer figures from telematics where you have it, books services at the intervals you set and reminds the fleet manager before each renewal. A machine that is overdue for inspection is flagged so it is not sent to site.
Stop typing the job in twice
A won job is entered in the estimating system, again in the job management system, and a third time in the accounting system when it is invoiced. Cost codes drift, and nobody trusts the job cost report. We treat all of that as one process and measure it that way. Nearly every step is rule-based, with mappings between the systems, a sync that runs on a schedule, a reconciliation that compares them and an exception queue for records that do not match. AI appears only where a free-text note has to be mapped to a cost code, and those mappings are reviewed before they are saved. The office ends up with one version of each job.
If one of these costs you real money every month, describe it in four answers.
Thinking about what a process actually costs you?
The assessment is small, fixed-price, and tells you the real number, whether or not you ever build.
Worth a 30-minute conversation.