The outcome
What changed
01 · Owner time
30 hrs
A week, returned to the businessUp to 10 hours each, across three working directors.
02 · Pricing memory
900+
Historical quotes made usableYears of spreadsheet pricing, consolidated into one costed source.
03 · Time to live
2 wks
From data to a working systemDelivered as a free proof of concept, tested against real enquiries.
04 · Defensibility
100%
Of priced lines cite their sourceWhere the data can't support a price, it says so.
Client named on request. Hours saved are the directors' own assessment of the manual work removed, not a modelled projection.
The problem
The quote
lived in
their heads
- Sector
- Branded merchandise and apparel
- Size
- SME, three working directors
- Region
- United Kingdom
- Process
- Inbound enquiry → manual costing → quotation
Every enquiry needed someone senior to price it — not because the task was hard, but because knowing which supplier, which quantity band and which of four near-identical product names sat with the people who'd been doing it for years.
That knowledge existed on paper too: more than 900 historical quotes across spreadsheets built by different people at different times, complete with duplicate product names and minimum order quantities that contradicted the quotes beside them.
So three directors spent their weeks reproducing decisions they'd already made. Quoting isn't the work that grows a merchandise business. Winning the account is.
The engagement
What we actually did
We didn't arrive with a product. We arrived with questions, and spent the first stretch working out how a quote genuinely gets made here — including the bits nobody had written down.
01
Mapped the real process
Understand before you build
Which decisions are judgement and which are lookup. We proposed three ways to take work out of their heads; they picked quoting, because it was eating the most of them.
02
Turned 900 quotes into one pricing memory
Data first, always
Historical sheets, price bands and product naming, consolidated into a single queryable source. Where rows disagreed we flagged them for a human rather than averaging them into a comfortable lie.
03
Built it into the tool they already use — email
No new software to learn
Forward an enquiry to an address; a costed answer comes back, itemised by line at the correct quantity band. No portal, no login, no change to how the sales team already works.
04
Made it show its working, and admit its limits
Traceable or it doesn't ship
Every priced line links to the row it came from, with the reasoning underneath. Where the data can't support a defensible number, it says so instead of inventing one. That's why the team trusts it.
How it works now
An enquiry in.
A defensible number out.
- An enquiry arrives as an ordinary forwarded email — the same way it always did
- Each line is costed at the right quantity band, not a flat rate
- The reply carries the total, the source row behind every price, and the reasoning in plain English
- Where the data can't support a number it is flagged, not guessed
- A director reviews and approves. Nothing reaches a customer unread
REQ-20260812-A41C · costed in 40 seconds
| Product | Qty | Cost/item | Line total |
|---|---|---|---|
| Embroidered beanie Sheet B, row 130 |
250 | £3.4800 | £870.00 |
| Heavyweight hoodie Sheet B, row 165 |
500 | £11.8500 | £5,925.00 |
| Organic cotton tee Sheet B, row 389 Cost understated · below MOQ of 250 |
120 | £4.9300 | £591.60 |
Illustrative reconstruction. Products, quantities and prices are invented for this page — no client pricing is shown.
The part most AI projects skip
Their data, on their terms
01
Your data never trains a model
It answers your questions and nothing else, under a zero-retention policy.
02
Read-only, scoped to what you hand over
We work from the slice of data you give us. The automation reads; it never writes back.
03
Credentials sit in a managed store
Not in code, not in spreadsheets, not in anyone's inbox.
04
A person approves anything customer-facing
The automation produces a draft. Someone at the client signs it off. That's the design, not a setting.
Honest assessment
Would this work
for your business?
The pattern travels well beyond merchandise — print, signage, fabrication, engineering, contract manufacturing. But it doesn't fit everyone.
Good fit
- Quoting depends on a handful of experienced people
- You have years of previous quotes, however messy
- An owner or director is doing work a system could do
- Pricing follows rules, even unwritten ones
Poor fit
- Every quote is genuinely bespoke with no precedent
- No historical pricing exists in any recoverable form
- Pricing is set live by negotiation, not by cost
- You want a number produced with nobody checking it
Questions we get asked
Quote automation, answered
Can AI generate quotes from our historical pricing data?
Yes, provided the history exists somewhere. Here, 900+ historical quotes sat across spreadsheets. We consolidated them into one costed source of truth, then built an automation that prices a new enquiry against it and shows which row it used. The answer is only as good as the history — which is why the data comes first, not the technology.
What if our quoting data is messy or inconsistent?
It usually is. Rather than smoothing over duplicate product names and conflicting minimum order quantities, the automation flags them: it prices what it can defend and says where the data will not support a number. Those flags become the improvement list for the business, and they're often worth more than the automation itself.
Does our data get used to train an AI model?
No. Client data answers that client's questions and nothing else. It is never used to train or fine-tune a model, it runs under a zero-retention policy, access is read-only and scoped to the data handed over, and credentials sit in a managed secret store.
How long does it take to automate quoting?
This went from data handover to a working system the sales team could use in two weeks, delivered as a free proof of concept. Timelines vary with the state of the data, but two to four weeks to a working first version is typical.
Does this only work for merchandise and apparel companies?
No. It applies wherever quoting depends on judgement held by a few experienced people plus a history of previous quotes: print, signage, fabrication, engineering, contract manufacturing and specialist distribution all fit. What matters is that past quotes exist, and that someone senior is spending hours a week reproducing them.
Quote automation · Don't Hire
WHAT
WOULD YOU
DO WITH
TEN HOURS?
- Ongoing engagements are fully managed — all usage and running costs included
- Up to 10 hours of our time each month, on improvements or the next process
- Fixed price, given after the first call
Fifteen minutes. Tell us how a quote gets made in your business and who currently makes it. We'll tell you straight whether this pattern applies, what it would take, and what it would cost. If it won't work, we'll say so on the call.
Book a free 15-min call →No obligation, no hard sell. We charge nothing for discovery.