The Fabrik Manifesto

“To find out what happens to a system when you interfere with it you have to interfere with it (not just passively observe it).”

— George E. P. Box, “Use and Abuse of Regression,” Technometrics, 1966

Since the industrial revolution, we’ve built machines to help us do more. A loom turned thread into cloth. Computers took on calculations and recordkeeping. Each became useful because people found work they could reliably entrust to it. We believe AI agents can earn a place alongside them.

There is plenty of work worth doing. Customers wait for answers while someone looks through records. Teams spend hours moving information between systems and following up on unfinished tasks. Agents could take on more of that work, giving people time for the decisions and relationships that need their attention.

A promising pilot is a start. Putting it into daily use asks more of everyone involved. Someone has to decide whether the agent can handle a customer’s request, whether it knows when to ask for help, and whether the people around it can depend on the result. The person making that decision will still be responsible when something goes wrong.

We think that hesitation deserves a serious answer. People need to see how an agent handles the work they intend to give it when a request leaves something out, records disagree, or a task stops halfway through. Confidence grows when a team can examine what happened, correct a mistake, and see whether the correction holds.

This should be possible before customers have to live with the mistakes. Teams need room to try things, find the limits, and improve an agent without putting the real work at risk. They should be able to explain why they are ready to move forward and where a person still needs to stay involved.

Trust also needs care after the first release. The work changes, the agent changes, and people find situations nobody anticipated. What a team learns should carry forward, making the next decision better informed. A successful test cannot answer every future question, but it can give people a firmer basis for the next step.

We started Fabrik to help make that step possible. We want the people building agents and the people responsible for their work to have a shared basis for trusting them. When they do, more promising ideas can make it out of pilots and into everyday use.

AI agents should earn our trust through the work they do.

Fabrik Labs