Inventum has been engineering heat pumps, boilers and hot water systems since 1908, and was first to bring a ventilation heat pump without an outdoor unit to the Dutch market, back in 1989. Today, around 100 people design and manufacture heat pumps across two factories in Houten. As the product range grows more complex, so does the pressure on the small group of people who hold the answers.
Knowledge that outpaces the people who hold it
Inventum has no shortage of technical knowledge: manuals, installation guides, service documentation, years of accumulated field experience. The problem was never that the knowledge didn't exist. It's that knowing where to find it, in the moment someone needs it, was never guaranteed.
Partners and installers already have a lot of information available to them on Inventum's website. But it regularly happens that either the answer isn't there, or people don't know it exists. So they call the service team instead, and that creates a queue. Internally, the same pattern repeats: recurring questions land back on the same few colleagues, and every email that goes to someone who can't quite answer it adds a delay before the customer gets what they need.
Inventum started looking for a way to make its own knowledge instantly findable. For its own team first, then for the wider partner network.
Choosing a dedicated partner over generic AI
Inventum already uses generic AI tools for various tasks. But for something this close to product safety and customer trust, generic wasn't good enough.
We wanted to safeguard quality, so we chose a partner dedicated to our segment, manufacturing. That, plus real support getting the implementation right, is why we chose Chapter.
Building trust before scaling up
Inventum's first phase wasn't about rolling out a tool to everyone. It was about proving, internally, that the answers could be trusted. That meant testing the system against real questions from real documentation, and watching how it behaved when a question was ambiguous.
The moment I saw this could really work was in the test phase, when I got answers with the sources right there: highlighted passages from our own documentation. When the system wasn't sure what I meant, it asked: 'Which product? What's your exact situation?' That's what builds trust in the answer.
That combination, a cited source plus a system that asks rather than guesses, is what moved the project from "interesting demo" to something the team is now using with real customers. Early on, the service and sales team already reached a point where they could give a verified answer directly at the table with a customer present, instead of promising to follow up later.
Early signals
The rollout starts with Inventum's own service and sales team before it extends outward. It's early, but the first uses are already pointing at exactly the kind of impact Inventum set out to capture:
One colleague put it simply: if I'd had this tool this morning, it would have saved me half an hour today. Often it's more than that. Sometimes it's the difference between waiting days for an answer and getting it right away.
That's the pattern Inventum is building toward at scale: less time spent searching, more time spent on the cases that actually need a person.
For Inventum, the case for Chapter was never just about speed. It's about what that speed frees up.
The value of Chapter for us is really threefold. It helps us bring customer satisfaction to a higher level, it frees up time for our team to focus on the more complex cases, and it gives us continuity. Every gap we close today means we can answer that question for good, tomorrow.
Each phase is designed to build on the last. Fewer routine questions reaching the service desk. New colleagues who no longer need weeks of shadowing a senior engineer before they can answer confidently on their own. A documentation base that keeps closing its own gaps as real questions surface them, instead of expertise staying locked in a handful of people's heads.
The numbers aren't in yet, but the shape of what's coming is already clear, and this case study will be updated with them as the rollout reaches installers and beyond.
