A 50-person contractor will never hire a full-time analyst
A 50-person contractor will never hire a full-time analyst. There is not $100k a year of analysis to do. So nobody checks which job types actually make money, which subs run late every time, or which estimator comes in 8% low on every bid.
That is the work AI makes affordable: a slice of a specialist the business was always too small to hire. Nobody can put 15% of an analyst on payroll. Something close to it now fits inside a software bill.
The catch is that some of that work went undone because it was never worth doing. A weekly 40-page competitor report costs almost nothing to produce now. Nobody will read it, and no bid will change because of it.
So before paying for any new capability, walk it through four questions. Which decision would it change? Why couldn't we make that decision well before? What would we do differently next Monday? What is that worth in margin or in jobs won?
If the chain breaks at any step, you have bought cheap paper.
If it holds, one analysis can outrun a year of hours saved. Learning that one job type loses money changes which jobs you bid, how you price them, who you send and which subs you call.