Most of the conversation about AI in this industry is about getting started. Which provider, what it costs, how long the build takes, what it will do on day one.
Very little of it is about month seven. Which is a shame, because month seven is where AI consulting for construction either proves it was worth doing or quietly falls over.
What month seven actually looks like
The build went well. Everyone was pleased. Then the business carried on being a business.
A supplier changed the layout of their paperwork, so the agent that was reading it started pulling the wrong field. The estimator who championed the whole thing has moved on, and the person who replaced them was never shown how any of it works. A new client came on with different requirements, and nobody adjusted anything to suit. The office has gone back to doing one part by hand, because it was quicker than working out why it stopped.
None of that is a failure of the technology. It is the ordinary business of a construction company changing faster than the thing that was built for it.
Why nobody asks about this up front
Buying decisions get made on a demonstration, and every demonstration is month one. Smooth, impressive, nothing has changed yet. The questions that come up are about capability, because capability is what is in front of you.
There is also no shared standard to measure anyone against. Australia's National AI Plan, released at the end of last year, stepped back from mandatory rules for high-risk AI in favour of existing law and guidance. Whatever you think of that as policy, the practical effect for a business owner is that nobody is being told what good looks like. There is no benchmark to hold a provider to, so people fall back on the demonstration.
Two things changed in Australia this year
The first is that the job management platforms this industry runs on have started adding their own AI features. Reading invoices, sorting emails, suggesting schedules. Useful, and worth knowing about. What they do is the generic middle of a workflow, the part that is the same for every customer of that platform. They are not built around how your business is different, they do not reach across two systems that were never meant to talk, and nobody adjusts them when your supplier changes their forms.
The second is that the number of providers offering AI consulting for construction has grown quickly. Some are very good. Plenty are two people and a portfolio page. Neither is obvious from the website, and the difference usually shows up around month seven, when somebody needs to answer the phone and make a change.
The questions worth asking before anyone builds anything
These are the ones worth putting to any provider, including us.
Who adjusts it when something upstream changes? Not whether support exists, but whether the person who takes the call has seen your build before.
What happens when a key person leaves? If the only person who understands it works in your office, that is a risk you have taken on, not one that has been handled.
Is it documented anywhere a new starter could follow? A lot of what gets built in this industry started as a favour from someone who was good with computers, and it was never written down.
What does the second year cost? If the answer is vague, that is the answer.
How we think about it
AI consulting for construction is not a project with an end date, and it is not honest to sell it as one. The build is the smallest part. Everything after it is where the value either holds or leaks away.
So the work is set up to continue. The same people who built it are the ones who adjust it. When a supplier changes their paperwork or a crew changes the way they work, it gets looked at rather than left to drift. When someone new joins the office, they get shown how it runs. That is not an add-on, it is the job.
It also runs on top of the systems already in place. Nothing gets replaced and nobody learns a new program, which matters enormously in month seven, because a business under pressure will always fall back on what it already knows. And the work that needs a person stays with your people. The client who needs handling carefully, the quote that is not standard, the call that needs judgement. Those come back to someone in the office, every time.
Start with one job, then decide
The best protection against the month seven problem is to keep the first build small. One repetitive job, usually the one costing the most, built properly and watched for a few months. If it holds up, there is a conversation about a second one. If it does not, you have learnt something useful without having rebuilt your business around it.
That is how AI consulting for construction should work for a business of twenty to fifty people with no IT department. Small, specific, and looked after.
If you are working out whether any of this is worth doing yet, that is a conversation we have most weeks. It does not cost anything, and we will tell you if the answer is not yet.