You told us the problem is the pipeline: too much of the work coming from too few contractors, and a need for more of it. Your own answers describe a business that is winning plenty. The margin leaks after you win. Price the same steel package through two of your estimators and it comes out about eight per cent apart, because each one prices off their own spreadsheet. Then jobs run over on the floor, and nobody sees the hit until the year-end accounts land.
The highest-value thing AI can do for Reynolds this year is make a job cost the same whoever prices it, and show you the margin while the job is still running rather than a year after it finished. Fixing what you keep is faster and cheaper than chasing more of what you win.
You are not short of work. You are short of a way to see the money leaving.
1. Buyers shortlist through AI before they call.
Main contractors and structural engineers are starting to scope steel packages and shortlist fabricators with an AI assistant before anyone picks up a phone. If the assistant doesn't name Reynolds when someone asks who fabricates structural steelwork in your region, you are not on the list, and you never find out you were dropped. Most fabricators have no presence there yet, which is exactly why it is winnable now.
2. Estimating stops being a craft and becomes a system.
The pricing judgement in an experienced estimator's head, how a connection detail drives labour hours, what steel-price movement does to a job priced today and delivered in three months, which packages are quietly unprofitable, is exactly what AI is now good at encoding. The firms that hold their margin through the next few years are the ones that captured that judgement instead of leaving it in two people's spreadsheets.
3. Job costing goes real-time.
Knowing a job's true margin while it is still on the floor, not at year-end, is about to become normal. That changes which work you chase and which you decline, because the thin jobs stop hiding inside a healthy-looking annual number.
Drawn from your Discovery Meeting. Single-interview engagement (MD only), so this is the founder's view, not a triangulated cross-role synthesis.
Your enquiry-to-despatch chain is full of repetitive, rules-based work, and some of it is a candidate for removal. But the Discovery Meeting was a strategic interview, not a stopwatch walk of the floor, so nobody asked how many hours a week each task eats. We won't invent the number. Half an hour on the floor fixes that.
One priority passed the test. We looked hard for a second and did not find one that met the bar. Padding this section would waste your attention.
Three moves, in the order they matter to Reynolds. Each one starts with something free you can do this week, then the AI that makes it stick.
Then the AI. An estimating assistant that reads the fabrication drawings, produces a first-pass take-off, and prices it against that single rate card, so the estimator checks a draft instead of starting from a blank sheet, and the number no longer depends on who picked up the enquiry.
Why it's first. It attacks the margin leak at the point the price is set. Every other move measures or reports the problem; this one removes its cause. It is also a genuine bet worth naming: nobody has written down how your estimators build a price, so the first month is capture and the second is build. We don't start building before we know what it has to do.
Then the AI. Margin by job and by month, the drivers underneath the numbers, a pack built automatically on a fixed date, and a live view of what you're owed, including which retentions fall due.
Why AI does this better than a person. The pack isn't late through incompetence. Assembling it is a manual job that always loses to whatever is on fire. AI produces it on the same date every month whether anyone chases it or not, and it tracks the drivers, so a job running thin surfaces while you can still act on it rather than a year later when you can only regret it.
Then the AI. It plans and writes content in your voice on a schedule, and runs AEO (Answer Engine Optimisation): making sure Reynolds is named when a contractor asks an assistant for a steel fabricator. The new version of showing up first on Google, except the buyer now asks an AI and acts on the two or three names it returns.
Why not first. AI visibility takes months to move, so the baseline starts now. But it does not fix this year's margin, and it shouldn't take your attention while the same job still prices two ways. The pipeline you named gets solved here, once the leaks above are closed.
This section would normally put a number of hours a week against each move. We can't do that yet, so we haven't.
Figures are directional and use Reynolds' own estimates where the Discovery Meeting captured them. Where hours could not be captured, we have said so rather than estimated.
Three things, in this order. One: write the rate card, a day with both estimators, no AI, and it removes most of the spread on its own. Two: give us half an hour on the floor about volumes, so the next version of this document puts real numbers where the gaps currently are. Three: build the AI Foundation, then the estimating assistant and the monthly job-costing view run on our side.
One interview (MD). Single-voice engagement, so no cross-checking against other roles was possible, and no cross-cutting themes have been invented from one account. Where Jane's diagnosis and her evidence disagree, we have surfaced the disagreement rather than resolved it. That is her call, and it is the first question we would like her to answer.
Reynolds & Co is invented. The numbers, the names and the situation are illustrative, chosen to show a typical SME shape: a sound operation with money leaking through processes nobody has had time to fix.
What is real is the format. Every Blueprint follows this structure: a straight verdict, an executive summary, where the industry is heading, what we found, why to act now, what to do in the order that matters to you, the expected impact, and a per-person view. It is built entirely from your own Discovery Meeting, and it is yours to keep whether or not you go further.
A Discovery Meeting takes 30 minutes. You walk away with a Blueprint like this one, built on your business, not a fictional one. rq@aigentic.co.uk