Four times the volume.
Same team.No new hires.
So we built a configurator
that cuts too.
A Belgian family manufacturer was handling service requests across five countries by hand 20 to 30 minutes each, with trained staff working through dozens of decision points. We turned the decision tree into software. configurator that assembles the product from cut profiles at runtime, produces the bill of materials, the quote, the order and the packing list.
- Reduction in processing time per request
Reduction in processing time per request
- Time to assess a single servicerequest
Time to assess a single service
request
- Volume handled, with the same team
Volume handled, with the same team
- In continuous daily use since deployment
In continuous daily use since deployment
Thirty years old, family run, market leader at home.
A Belgian family business with more than thirty years in small electronics deep fryers, air fryers, electric barbecues, waffle makers. Around 146 products, sold across five European markets, with a reputation built on quality and design rather than price.Exactly the kind of company this work suits: successful enough to have outgrown manual processes, lean enough that the people running them have other jobs to do.
Five countries, five rulebooks, one trained person at a time.
Every service request had to be assessed by hand. Is the product in warranty? Which warranty the terms differ by country. Is this model repairable, replaceable, or end-of-life? Which service partner covers this postcode? What does the customer get told, and in which language?
A trained member of staff worked through those decision points for 20 to 30 minutes per request. The logic wasn't written down anywhere complete. It lived in the heads of the people who had been doing it longest.
Hundreds of possible outcomes.
Warranty status, product type, purchase date, country and service partner combined into a decision tree no one could hold reliably in their head. Mistakes were inevitable, and each one meant a customer told the wrong thing.
Five sets of rules that kept diverging.
Belgium, Germany, France, Luxembourg and the Netherlands each had different service policies and warranty terms. Every policy change had to be communicated to every person handling requests, and remembered.
Growth meant hiring, or waiting.
Rising volumes left two options: add trained staff, or let response times slide. For a company competing on quality of service, neither was acceptable.
The knowledge existed. It just wasn't anywhere a computer could reach it or anywhere a new employee could learn it in less than months.
Document the logic first. Build second.
The build wasn't the hard part. Extracting the rules was.
We sat with the people doing the assessments and worked through real requests including the awkward ones, the exceptions, and the cases where the answer was "it depends who's asking." Every branch of the decision tree was documented and signed off before a line of code was written.
That document turned out to be valuable in its own right. It was the first time the company's service logic existed anywhere outside the memory of the staff running it.
- Request in
- Warranty check
- Country rules
- Product status
- Routing
- Answer in the customer's language
A decision engine, and a form simple enough that nobody needed training.
Automated warranty assessment.
Purchase date, product and country resolve to a warranty position without anyone looking anything up.
Country-specific routing.
Each market's policies encoded separately, so a change in Dutch terms doesn't touch the German ones. Policy updates are a configuration change, not a retraining exercise.
Multi-language throughout.
Staff and customers each see their own language. In a five-country operation this isn't a nicety it's the difference between a system people use and one they work around.
Exception handling that escalates rather than blocks.
Cases the rules can't resolve go to a person, flagged, with the reasoning visible. The system never guesses, and never silently gets it wrong.
Visibility and reporting.
Complete request history, live status, and analytics on what's actually failing in the field which products, which markets, which faults.
What changed
Assessment time fell from 20–30 minutes to under two.
A 95% reduction. The work that took a trained person half an hour now takes a form and a moment.
Volume grew fourfold on the same headcount.
Growth stopped being a staffing question. The team that handled the old volume handles four times as much.
99.8% accuracy on recommendations.
The errors that came from holding a five-country decision tree in your head largely disappeared.
Service available around the clock, in five markets.
Requests no longer wait for the right person to be at their desk in the right country.
New staff productive in a day, not months.
The expertise sits in the system. Onboarding stopped being an apprenticeship.
That last detail wasn't in any specification. It came from listening to what actually worries someone receiving forty boxes of aluminium profile.
We deployed it in 2017. It's still running.
Not rebuilt. Not migrated to something else. Not quietly abandoned after the people who commissioned it moved on. The same system, handling service requests across five European markets, every working day since late 2017.
We are not involved in running it, and haven't been for years. There is no licence to renew, no annual invoice, no dependency on us continuing to exist. It was built, handed over, and it works.
That's the standard we build to. A system that needs its author nearby to keep functioning isn't finished it's a subscription with extra steps.
The best thing we can say about a system is that the client stopped needing to think about us.
It's the same problem as quoting.
This wasn't a configurator. But it was the same shape of problem, and it's worth naming why. A trained person navigating hundreds of possible outcomes, applying rules that differ by market, arriving at an answer nobody else in the company could reproduce that is precisely what happens when a manufacturer quotes a custom product. The variables differ. The structure doesn't. If the knowledge for something in your business lives in one or two people's heads, and everyone waits on them, it can usually be built into software. Quoting, estimation, specification, service, compliance. The question is never the industry. It's whether the logic can be written down.
Have a process that only two people understand?
Thirty minutes. Walk us through one real case the messy kind, with the exceptions. We'll tell you honestly whether it can be automated, roughly what it would cost, and how long it would take.