Quoting and procurement automation in UK manufacturing
Quoting and procurement look like different departments, but in most UK manufacturers they are the same loop run in two directions. A customer sends drawings and a specification; someone extracts what is actually being asked for, looks up what it cost last time, chases suppliers for the prices they do not have, and assembles a number before the customer loses patience. Then, when the job lands, the same firm becomes the customer: sending its own enquiries out, waiting on its own suppliers, chasing its own quotes.
Both directions are extraction plus lookup plus follow-up, and all three of those are now automatable to a useful standard. What is not automatable is the engineering judgement in the middle, and any vendor telling you otherwise is selling the demo, not the tool.
Why do quotes take two weeks when the thinking takes two hours?
Because the thinking is buffered on both sides by waiting.
Before the estimator can think, someone has to turn a zip file of PDFs into a list of parts, materials, quantities, tolerances and finishes. That is reading and transcription, it takes hours, and it queues behind live jobs. After the estimator has thought, the quote waits on supplier prices: three emails sent, one answered, and nobody's diary says "chase the other two on Thursday". The enquiry that arrived on the 1st goes out priced on the 12th, and the firm that answered on the 3rd is already in contract. Make UK has long pointed to exactly this sort of process digitisation as where the sector's productivity gains are hiding, and quoting is a purer example than most factory floors.
So the automation target is not the estimator's judgement. It is the transcription before it and the chasing after it.
What is realistic now, and what is still oversold?
This is the table to argue with vendors over.
| Capability | Realistic today? | Notes |
|---|---|---|
| Structured extraction from drawings and specs | Yes | Pulling part lists, materials, quantities, finishes and notes into a structured sheet for review |
| Historical price recall | Yes | "We quoted a near-identical bracket in March, here is the price and the margin we got" |
| Supplier chase sequences | Yes | Enquiries logged, reminders sent on a schedule, responses filed against the enquiry |
| Quote assembly from approved rates | Yes, with review | Drafting the document from rates a person has signed off |
| Choosing the machining strategy or process route | No | This is engineering judgement; the tool should propose nothing here without a person |
| Fully automated quoting, no human review | No | Anyone offering this for non-trivial work is quoting your risk appetite, not your parts |
The honest line runs between recall and judgement. Software is genuinely good at "find every time we made something like this and show me the numbers", and that alone changes the estimator's morning. It is not good at knowing that this alloy in this tolerance band will fight the machinist, and standards compliance matters here too: where a specification calls up a standard (from BSI, typically), a person confirms what complying costs, because the consequences of guessing are not symmetrical.
What does the procurement half look like?
Like sales, but with the shoe on the other foot, which is a useful discipline. Your outbound enquiry is somebody else's inbox problem, so the wins are the same ones in reverse.
An enquiry that goes out as structured data (part, quantity, material, date needed) gets answered faster than a forwarded email chain. A chase sequence that follows up on day 3 and day 7 without anyone remembering to do it turns "one reply out of three" into three replies, and three replies is a price comparison instead of a hostage situation. And supplier responses filed against the enquiry, rather than scattered through a shared inbox, mean the buyer can see coverage at a glance: which lines are priced, which are stale, which are silent.
None of this needs new judgement. It needs the follow-up to happen every time, which is the one thing busy humans reliably fail at and machines reliably do not. That is the case made in zero-click automation, and procurement chasing may be its cleanest example anywhere in a manufacturing business.
How do you trial this without disrupting live quoting?
Run it alongside, not instead. Pick one estimator and one family of parts. For a month, every enquiry in that family gets the automated treatment in parallel: extraction into a structured sheet, historical matches surfaced, supplier chases sent by the system. The estimator still quotes exactly as before, but starts from the sheet instead of the PDF pile.
Measure two things: hours from enquiry received to quote sent, and how often the historical match was close enough to be useful. Those two numbers tell you whether to expand, and they are yours, not a vendor's. This is the discipline described in what a good AI pilot looks like: small scope, real work, numbers agreed before you start.
What you are likely to find is that the bottleneck was never the estimating brain. It was the reading before it and the chasing after it, and both of those are now a solved problem for firms willing to spend a month proving it. AI Metric builds this sort of document automation for exactly this trial-first approach, because the loop is the same in every firm and the proof has to be local.
Quote faster than your competitors and you will win work you currently never see, because the jobs are being placed before your number arrives.