Trackunit · Platform & AI
A long tail of requests. Let customers assemble the answer.
Instead of building one feature per request, we opened Access Management and Inspections to AI through MCP connectors, so customers could build their own reports and workflows with the assistants they already use.
The problem
Each request reasonable. Together impossible.
Customers had a long tail of wishes: custom reports, bulk actions, configurations, small automations. Each one made sense on its own.
The classic route is one feature per request — until the product is a pile of special cases, and customers still do the rest by hand.
My responsibility
I owned the decision to make Access Management and Inspections composable by AI, and the boundary for what AI may do on its own and where a person stays in control. The connectors were built by the team, not by me — my part was the product call and the line we drew.
“…the first backend team at Trackunit to deliver our services as MCP tools.”
The decision that mattered
Make the data addressable — not every screen.
MCP connectors across the access management and inspections data, so customers could assemble reports, bulk actions and automated workflows themselves, through assistants such as Claude. Launched as a deliberately bounded MVP: in access management and safety, the failure mode of a wrong bulk action is the wrong person on a machine — so the boundary was a safety decision, not a technicality.
What happened
The connectors launched, and customers could build what we would never have prioritised one request at a time. The rest was meant to be learned from real use rather than designed up front.
“Anders moved the product discussion from features to solutions for customer problems.”
What I learned
- Composability is a product choice with a price: you give up control over what users build, and in return you get a long tail you never have to build yourself.
- A list of things the model must not touch is out of date the moment the product changes. Boundaries hold better when they are built into the product than when they are written as rules.
- “MVP” is a stronger word when you say it yourself. A team that has shipped and is learning beats a team with a governance model no user has met.