Eighty per cent of building businesses that went all-in on AI have since walked it back. What that reversal tells us about how AI actually fits in a building business.
There was a moment, not long ago, when AI felt like it was going to solve everything.
Estimating. Marketing. Cold outreach. Answering the phone. Writing the social calendar. Running the inbox.
Builders across Australia tried it. Some went all the way in. And then, quietly, most of them started pulling back.
That is the pattern Gavin Sloan, founder of Live Chat Monitoring, is seeing across the building businesses his platform serves. His data suggests around 80 per cent of businesses that fully committed to AI for key functions have since reversed course, bringing those tasks back in-house or hiring people to work alongside the technology.
It is not a story about AI failing. It is a story about what AI actually is, and what it is not.
The Honeymoon Phase
Aaron Ng, founder of The Good Builder, has been through this cycle himself.
“When we started The Good Builder, I went through a phase where I went full AI. We were getting AI to write our articles, getting AI to answer my emails, getting AI to do cold outreach.”
The cracks appeared gradually. The outreach emails started including services the business had never offered. Then the hallucinated services expanded across subsequent messages, building on themselves. The content felt flat. The voice disappeared.
“Everyone used to go, what’s up with this? That’s really weird. It doesn’t sound like us.”
Ng pulled everything back. Today, AI is a tool in the workflow. It does not run it.
“AI doesn’t know how your business works. It doesn’t know how customers respond to you. It doesn’t know the exact specifications of what your customers want. Only you can do that.”
The Disconnect Nobody Is Talking About
Nayan Kavishka, who works with builders across Australia through Scale Up Smart, puts the issue clearly.
The problem is not that AI is incapable. The problem is how it is being sold to builders.
“You’ve got software companies, even some coaches, who say you’ve got to use AI, you’ve got to do this, because if you don’t, you’re going to go broke. And what happens is builders download tools, spend the whole week trying to figure out a workflow, but AI doesn’t know how your business works. It doesn’t know how customers respond to you. It cannot replace the relationship.”
Builders who got excited, downloaded tools, and expected results quickly found themselves more distracted than productive. The technology made it easy to explore possibilities that had nothing to do with the actual business. New workflows got started and abandoned. Time disappeared.
“There is no system underneath,” says Anisha from Scale Up Smart. “There is no person actually checking the output.”
The Tool Analogy That Gets It Right
Anisha frames AI the way tradespeople understand tools best.
“A laser level doesn’t lay the tiles. A nail gun doesn’t frame the house. You still need someone who knows how to do the proper work.”
AI is no different. It can move faster, help you start something, surface a structure, or summarise a meeting. But without the person who understands the job, the client, and the context, the output lands wrong.
That is especially true in construction, where no two jobs are the same. Ng points out that building is not a binary industry.
“You never turn up to a site and it’s one plus one equals two and we’re out of here. There’s always grey.”
AI is very good at black and white. Construction rarely offers it any.
Where It Is Actually Working
The builders using AI well are not trying to replace judgment. They are trying to reduce friction in the parts of the day that drain time without requiring it.
Ng summarises his own use case simply: “I get it to summarise my emails every morning and arrange them from urgent to non-urgent. That clarity for the day, knowing what I’ve got on and what I missed, that genuinely helps.”
He does not let AI write the emails. He writes them himself.
Kavishka describes a similar pattern at Scale Up Smart. After meetings that run an hour or two, AI is used to extract the key points, create action summaries, and prepare the agenda for the next session. A human reads it, adjusts the tone, and makes sure it matches the client relationship.
“You get the notes, ask it to summarise and give you the important points. It has made things a lot easier for overcoming time-consuming generic tasks.”
Estimators at Scale Up Smart also use AI as a double-check layer, not as the primary calculator. It flags potential errors. A human still signs off.
“AI is good. But as a tool, not a team. It’s not a substitute for people.”
The Builder Who Got It Right
One builder Ng spoke with recently stands out for having worked through the full cycle and arrived somewhere practical.
This builder rejected AI for most tasks. He found it produced worse results than people for anything involving relationships or judgment. But he kept it for one specific function: giving him a second set of eyes on the business as a whole.
Financials. Trade scheduling. Overall status. Not to make decisions, but to confirm he was on track before he made them himself.
“That’s the penny drop moment,” says Kavishka. “Use it to support your work, not to finish it or replace it.”
This is the version of AI that survives contact with reality in a building business. It is narrow, specific, and supervised. It handles the tasks that consume time without requiring the knowledge and relationships that actually build a business.
The Human Detection Problem
There is another layer to this that builders increasingly feel but do not always name.
Clients are getting better at spotting AI-generated content. The cadence is wrong. The warmth is absent. The specificity is missing.
“People feel it’s lazy,” says Ng. “They feel you haven’t put in the effort.”
That perception matters most in a relationship-based industry. When a client is spending hundreds of thousands of dollars with a builder, they want to know they are dealing with a human being who understands their project. An automated response, however polished, works against that.
Ng tried an AI voice answering calls. The technology handled the mechanical part fine. But the conversation had no emotional intelligence. It wanted a name and a number and then it was done.
“I felt like, you have no interest in actually talking to me.”
Construction is built on trust. Trust is built through people. That sequence does not change because a tool exists that can mimic parts of the interaction.
The Warning Worth Heeding
There is a longer-term question sitting underneath all of this that does not have a clean answer yet.
Roughly 70 per cent of new content on the internet is now estimated to be AI-generated. A significant proportion of it contains errors. AI systems train on available data, including the content they have already generated.
Sloan raised this concern directly with Ng. If systems feed on their own inaccurate output, the reliability of AI-sourced information degrades over time. The confident answers become less trustworthy even as the technology becomes more capable.
For an industry where specifications, compliance requirements, and cost estimates carry real-world consequences, that degradation matters.
This is not an argument against using AI. It is an argument for maintaining human judgment at the point where outputs actually count.
The Good Builder Take
The builders who have found a workable relationship with AI have one thing in common. They treat it like any other tool: useful when used for the right job, a liability when handed tasks that require something it cannot provide.
The ones who went all in and came back report the same experience. The technology took over tasks it was not equipped to handle. Voice disappeared. Relationships suffered. Productivity, in some cases, actually fell.
The reset did not mean abandoning AI. It meant defining exactly where it earns its place.
Summarise the emails. Check the schedule. Flag the errors. Take the meeting notes.
Write the proposal? Call the client? Build the relationship? That still needs a person.
Get that distinction right, and AI becomes genuinely useful. Get it wrong, and you spend a lot of time and money going in a circle.
More AI news: Your AI Is Loyal to Everyone. That’s the Problem. Why Guardrails are so Important
General information only. This article does not constitute business, legal or financial advice. Always seek professional guidance relevant to your specific circumstances.
0 Comments