The report ran four pages. An engineering firm I work with had put its won and lost bids through an AI analysis, and three lines in it mattered more than the rest. On bids under $1,000, the firm won nearly every time. Under $5,000, it won about half. Under $10,000, about a quarter.
Read quickly, that looks like good news about small jobs: the firm almost never loses them, so go get more of them. The finding that changed the conversation sat underneath the win rates. A $1,000 bid takes as much of the team’s time to put together as a large one, sometimes more. The firm had been spending some of its most expensive hours on the work worth the least, and nobody inside it had lined those two facts up, because nobody had a reason to.
Most of what owners hear about AI is about writing more, faster: emails and social posts. I do plenty of that myself. When people ask me what I use it for, I answer with a question: “What don’t I use AI for?” But I think the most valuable thing it does for a small business is read. Point it at what the company has already done, and it can show the owner something that has been sitting in the files for years. What happens after the answer comes back belongs to the owner, and most owners aren’t expecting that part.
You have to be very careful and discerning of the response that it gives you.
— Gary Henson, Consultant
The report was only as good as the question behind it, and that is true of every AI tool I use. “Garbage in, garbage out.” I rarely take the first answer. I keep the dialogue going, asking again and narrowing, until what comes back holds up against what I know about the business. Used that way, it saves me somewhere between 1 and 10 hours on a project.
Asking the right question is the same skill I use with a prospect. I keep asking until the owner says something true about the business, and it usually takes more than one try. An owner who hasn’t decided what he’s looking for won’t get much out of an AI tool pointed at his business, either. A loose request gets you a tidy summary. A pointed one, such as where the team’s hours are going compared with what each kind of job is worth, gets you something you can act on.
Then the answer needs a careful read. “You have to be very careful and discerning of the response that it gives you.” The tool knows the numbers you handed it. It doesn’t know which small job came from a customer you’ve served for 20 years, or which one opened the door to a much bigger contract the next spring. That judgment belongs to the people who were there, and it’s why I tell owners, “AI isn’t the end-all, be-all for every person and every circumstance in every business.”
This is where AI turns into a leadership question. A report like that is easy to read alone and file away. It’s harder to sit down with the team and say what it implies: some of the work we’ve been glad to win has been costing us. An owner who leans toward distrust will read it looking for someone to blame. I’d rather see it taken to the people who prepare the bids and put to them plainly: what’s working, what’s not working, and what’s missing in how we decide which jobs to chase?
Those hours belong to people. Every hour spent putting together a bid the firm rarely profits from is an hour someone on the team could have spent on work that grows the company, or on getting home on time. The employees are the internal customer, and the measure I use for any AI tool is how much of their time it gives back.
I don’t know yet how this one ends. The firm is using the report to rethink its pricing and decide where the team’s time goes, and that work is still underway. It connects to a question I’ve written about before, who pays when you don’t raise your prices, and to how narrowly most teams use these tools, which I covered in Most Teams Use AI for One or Two Tasks and Stop There. Taking a finding like this to the whole company, and holding everyone to what gets decided, is part of the Inside-Out Method.
Somewhere in your business there is a record like those bids: quotes, job costs, overtime, returns, the calls that never turned into work. You probably already know the question you’d be a little uneasy asking about it. Before you ask, decide who will look at the answer with you, and whether you’re prepared to change something if it says what you were afraid it would.


