Case study 01 · Industrial distribution, Europe
Good business, old systems.
A European industrial group — profitable, growing, and quietly capped by how much its people could physically get through in a day. We put an AI operating layer inside it, starting with the customer inbox and working down the whole order cycle behind it.
The company
Everything a good business is supposed to be.
A family-owned European group selling and servicing industrial components through a network of branches and field service teams across twenty-plus countries. Their customers are industrial and infrastructure operators — businesses where a stopped machine costs money by the hour, and where the supplier who answers first usually wins the order.
Long-standing repeat customers. Real margin. Real assets. Nothing wrong with the business at all.
What ran it was a shared mailbox, a telephone, a spreadsheet, and two different systems of record inherited from two different eras. A customer would write in describing a part rather than quoting a code. Someone would recognise the description, look the item up among several hundred thousand, check what was actually on the shelf, find that account's negotiated price, work out whether it could still make today's dispatch, and write back. Several hundred times a day, in several languages.
One request, start to finish
What actually happens between arrival and reply.
Taken from the system running today. The example happens to be an industrial customer asking about a part. It could as easily be a patient enquiry, an insurance claim or a freight booking — the shape does not change.
Understands the ask, and who is asking
Three questions in one message. The sender is matched to their account, and with it their contract terms, their location, their credit position and twelve years of order history.
Finds the real answers
A described product, not a code — narrowed out of several hundred thousand by meaning, to the item this account has bought eleven times before. Then live stock, their contracted price, today's dispatch cut-off. Every fact from the company's own systems. Nothing estimated.
Decides whether it is certain enough
The part that matters. Clean answers from live data, and it proceeds. Ambiguous description, stock split across locations, pricing off-contract — it stops and hands the email to the right person with all the research already done. It does not guess at a price or a date.
Replies, and learns
In the customer's language and the company's own voice, every figure shown with the source it came from. A person reads it and clicks send. Whatever they change feeds straight back in.
Trust, and who holds the dial
The line moves — and the client is the one who moves it.
On day one it drafts and the team approves. That is deliberate: it is how the system earns trust, and how their people teach it. As accuracy on a given category proves itself month after month, the client releases that category to go out on its own.
Simple stock and order-status answers usually go first. Pricing and technical recommendations stay with a person far longer, sometimes permanently. Over time most of the routine volume ends up fully automatic — but every step of that is the client's decision, on their own evidence, category by category. We never move that line for them.
Anything that constitutes a commitment — a price, a stock figure, a delivery date, a warranty term — can only ever come from data retrieved out of their own systems. Where the system cannot find the answer, it routes to a person rather than filling the gap.
What it delivers
The same team covers far more ground.
The business gets faster at exactly the moments that decide whether it wins the order.
| Unit economics | Before | After |
|---|---|---|
| Routine requests answered without a person drafting | None | 80%+ |
| Quote turnaround | Same day, at best | Under an hour |
| Quotes that get followed up | When someone remembers | Every one, on schedule |
| Orders needing manual entry | Nearly all of them | Under 10% |
| Supplier invoices touched by a person | Every one | Only the disputes |
| Days to get paid | Whenever someone chases | Weeks earlier |
| Revenue per employee | Flat for years | Up 80–100% |
| Gross profit per employee | Flat for years | Up 80–100% |
| Operating profit margin | Around 10% | Around 30% |
The same person who spent the morning re-typing purchase orders spends it selling, quoting and solving the technical problems that win accounts. They are individually two to three times as valuable to the business — and because the system carries the company's best practice, a two-year employee answers with the knowledge of a twenty-year one. A team that could only quote what it had hours for now quotes everything, chases everything, and answers everyone the same day.
That is why the revenue line and the cost line move at the same time. It is also why this tends to be popular with the team rather than resisted. Nobody's job was ever to re-key a document.
The email is only where it enters
The value compounds the further down the cycle it goes.
That enquiry is the front door of a cycle which today gets carried by hand from one system to the next, losing days at every hop.
- EnquiryArrives by email, phone or portal. Understood, attributed to the right account, routed or answered.
- QuotePriced off the customer's real contract terms and live availability, drafted in full, out the same hour instead of the next day.
- Follow-upQuotes that go quiet get chased on schedule. This alone recovers business that was already won and then forgotten.
- OrderThe customer says yes, or sends a purchase order as an attachment. Read, matched line by line, written straight into the system of record.
- FulfilmentStock allocated, shortfalls caught early, orders raised to the right suppliers. If a date moves, the customer hears it before they have to ask.
- InvoiceYour invoice raised off what actually shipped. Supplier invoices matched against order and receipt, with only genuine discrepancies reaching a person.
- CashEvery open receivable tracked and followed up politely, on time, every week — so money arrives weeks earlier.
- Next orderThe account's rhythm is learned. What they will need next is anticipated, and the cycle starts again warm instead of cold.
Scale & delivery
One function proves itself in the numbers, then the next one goes on.
The operating layer goes in at customer service and works its way down the order cycle behind it: enquiry, quote, follow-up, order entry into both systems of record, purchasing and stock, the counter and the service operation. Each function is scoped on its own business case rather than sold as one enormous programme. Nothing is committed on the strength of a slide — it is committed on the strength of the previous function's numbers. The same layer is rolling out across the group's twenty-plus countries.
How it is built: entirely inside the client's own cloud, in their own region. Nothing leaves. No copy on our side, under a signed data protection agreement, with named and revocable access and audit logs held on the client's side.
Client name available under a non-disclosure agreement.