Joost van Bruggen, MavenBlue
How an insurtech whose customers check every vendor's data handling puts AI to work every day, in under three minutes.
Let’s talk MavenBlue builds software for actuaries at insurance companies, and their customers want proof of how every vendor handles their data. Joost van Bruggen took AI on for the company and started out a sceptic. He tells how MavenBlue now runs AI on its own rules, on the AI it chooses, with its customers' data safe.
MavenBlue, a Dutch insurtech in Utrecht. Software for pricing, reserving and capital management at insurance companies.
Their customers’ data and personal data, under their own classification from level one to level five. ISO 27001 and SOC 2 Type 2.
Answering RFPs, understanding customers’ Excel files, asking questions of their own documentation, checking invoices, and documents in their own templates.
MavenBlue sorts everything it holds into five levels. Levels one and two are generic and can be shared. Level three is MavenBlue’s own information. Level four is customer data, level five is personal data. Joost wrote the rule for AI on top of it: levels four and five only go through Aimable, on AI that MavenBlue chooses and runs under its own rules.
That rule came first, and Aimable is how their people use AI within it: on a customer’s complicated Excel model, on an RFP that quotes from earlier ones, and on the functional and technical documentation that used to take twenty documents and a lot of searching.

“I think it took us, let’s say, two weeks before, and now we can do it in a day.”
MavenBlue gets more RFPs every year, most of them from outside the Netherlands and in English, and each one used to take about two weeks. They first put every RFP they had already answered into Aimable, together with the documentation behind it. Aimable built the RFP skill for them; the next skills, for documents in their own templates, Joost builds himself.
A new RFP now reuses what they wrote before, so it gets better with every round. The answers are more consistent across questions and across RFPs, grounded in MavenBlue’s own information, and in proper English for a team that does not speak it natively.
Aimable sits between MavenBlue’s own information and the AI outside. Their spaces, documents and skills stay as they are, while they try different AI per task: one that is good at programming for the Excel models, another for the RFPs.
Switching takes Joost about a minute. MavenBlue is not tied to one provider, which matters to a company whose customers check every vendor, and in a field that changes this fast.
“If you use AI in the right way, it can really help you. If you don’t use it in the right way, it can really hurt you. It’s like a magnifier of your own capabilities.”
How an insurtech whose customers check every vendor's data handling puts AI to work every day, in under three minutes.
How a 150-year-old candle company got from 35 possible use cases to the first three working AI apps.
Watch the webinar Five things an investment team needs to get right when AI joins the deal process.
Why HR2day put AI to work on the most sensitive data there is, and what it needed to trust it.
Investment memos, support answers, tenders, order intake: every customer began with one process, and none of them knew which one on day one. We help you find yours.
Shall we find yours together?