Written for builders, contractors, service businesses and distributors. Figures current at August 2026.
In the 2024-25 Business Characteristics Survey, 14% of wholesale trade businesses reported using AI. Construction was 6%. Transport, postal and warehousing was 1%, the lowest of any industry measured, against around 12 per cent of Australian businesses overall.
Two readings of that are available and only one of them is right. The wrong one is that these industries are behind and need to catch up. The right one is that most of the AI products sold in the last two years were built for people who sit at a desk all day and type, and a business whose value is created in a van, on a site or in a warehouse was correct to ignore them.
There is still real money here. It sits in a narrower band than it does in an office business, and it sits almost entirely in the paperwork that surrounds the physical work rather than in the work itself.
The three that pay
The gap between the site and the quote. Someone walks a job, describes what is needed into their phone, and that becomes a structured draft quote against your own price list before they get back in the vehicle. The value is not the writing. It is the number of days between the visit and the quote landing, which in most of these businesses is somewhere between three and never, and which is the single biggest reason work is lost to whoever answered first.
Answering questions from your own records. What did we quote this customer last time, what is the code for that fitting, which supplier had the shorter lead time, what did we agree on the variation. This is currently done by interrupting the one person who remembers, and that person is usually the owner or the ops manager.
After hours enquiries. Not a chatbot pretending to be a person. A triage that captures what the caller needs while it is fresh, sorts urgent from routine, and puts it in front of a human at seven the next morning with the detail already written down.
The two that do not
Pricing. The margin call on a job depends on things that are not written down anywhere: how that client behaves when something goes wrong, what the crew is like on that kind of work, what else is on next month. A tool that never sees any of that will produce a confident number and it will be wrong in a direction you only find out about at the end.
Anything a regulator or an insurer reads. Safety documentation, incident reports, compliance certificates. The failure mode here is not a bad first draft. It is that the document reads as competent, gets signed, and nobody notices what is missing until it matters.
The constraint that decides this
Whether any of it works comes down to one thing, and it is not the tool. It is whether your records are good enough to answer from. If the price list is current, the job history is in one system and the customer file is where it should be, all three of the workflows above are straightforward. If there are four versions of the price list and half the job history is in a text thread on somebody's phone, the honest first project is fixing the records, and no software will do that part for you.
That is a real finding rather than a disappointing one. It is also cheaper than the alternative, which is paying for licences that sit on top of records nobody trusts.
Start with one crew or one branch, measure how long the task takes before you change it, and give it to the person who already owns the process rather than to whoever is most interested in technology.