The Sydney Opera House was supposed to take five years. It took fourteen.
It was budgeted at seven million dollars. It finished at one hundred and two million. Around 1,357 per cent over budget in real terms.
The architect quit, left the country, and never came back to see it finished.
We put it on postcards. The rest of the world put it in textbooks, as the most famous example of what happens when you start building before you finish thinking.
Now bring that closer to home.
One in Two Hundred
A professor at Oxford named Bent Flyvbjerg spent decades building a database of more than 16,000 projects. Roads, rail, homes, renovations, stadiums, IT. He put the findings in a book called How Big Things Get Done, written with Dan Gardner and published in 2023, and the numbers should stop every builder cold.
Fewer than half of all projects come in on budget.
Eight and a half per cent come in on budget and on time.
On budget, on time, and delivering what they promised? Half of one per cent.
One in two hundred.
And his conclusion, after all that data, comes down to four words. Think slow, act fast.
Projects do not fail on site. They fail in the office, months before the slab. Planning is cheap. Construction is expensive. A mistake found on paper costs you an hour. The same mistake found on site costs you a variation, a fight and a month.
Flyvbjerg even has a name for what long jobs are exposed to. The window of doom. Every extra month on site is another month the window stays open. Weather. Price rises. A subbie going broke. A client changing their mind. The fast jobs are not lucky. They were planned so hard there was nothing left to go wrong slowly.
The Builder’s Version of All This
The job that hurt you this year was not lost on site. It was lost at quote stage.
A quantity taken off in a rush. An allowance that was really a guess. A PC sum covering the thing nobody wanted to pin down. A retaining wall sitting in the drawings but missing from the price.
Gaps.
Gaps become variations. Variations become arguments. Arguments become margin you never collect.
For twenty years the only defence was more hours, or a second estimator most builders could not afford. That excuse is gone.
Load the Whole Job Into One AI Project
The drawings. The spec. The scope. The quantities, the costings, the engineer’s details, the client brief. Everything. Then interrogate it.
What is in the drawings but not in my quote?
Where do my quantities not match the plans?
Which allowances look light for a job like this?
What do jobs like this normally include that mine does not?
That last question is Flyvbjerg’s biggest idea. He calls it reference class forecasting. In plain English, your project is not as unique as you think. Every builder tells me every job is different. Every job feels different. The ways jobs lose money are boringly the same. AI can hold your job up against the pattern and show you where yours is thin.
And it does in an hour what a second estimator did in a week. It does not get tired at page sixty of the spec. It does not skip the geotech report because it is Friday afternoon.
One Boundary, Stated Plainly
AI does not replace your judgment. It will get things wrong, and you check its work the same way you would check a first-year estimator’s. It is not there to price the job. It is there to make sure nothing is missing from the picture before you price it. You still make every call. You just make them with the whole job in front of you.
The Opera House did not fail for lack of talent. Utzon was one of the most gifted architects of his generation. It failed because construction started before the thinking finished.
Think slow. Act fast.
And for the first time in the history of this industry, thinking slow is fast.
Before you sign your next contract, load everything into one AI project and ask it one question. What have I missed?
The answer is worth more than the quote.
Common Questions
Can AI price a building job?
No, and you should not ask it to. AI is there to find what is missing from the picture before you price it. The pricing, the rates and the risk calls stay with you.
Will AI replace my estimator?
No. It does the checking a second estimator would do if you could afford one. It reads every page of the spec without getting tired. Your estimator still makes the decisions.
What should I actually load in?
Everything you have on the job. Drawings, specification, scope, quantities, costings, engineer’s details, geotech, the client brief. The value comes from AI seeing the whole job at once, which is the one thing a human reading page by page cannot do.
How accurate is it?
Accurate enough to be worth an hour, not accurate enough to trust unchecked. Treat every answer the way you would treat a first-year estimator’s work. Useful, and always verified.
If you want to see this done properly on a real job, book a free strategy session and we’ll walk through it together.
Kurt Hegetschweiler is the founder of Builders Coach and author of the internationally best-selling Million Dollar Builder. He has coached thousands of residential builders since 2004.
