“Good Enough”: A Field Guide to Quality in AI Agents

Kinomodus Slå av kinomodus Minispiller

Recorded 9 October 2026

What Does "Good Enough" Look Like? A quality workflow in agent-based applications

Marco Bertani-Økland, partner at Everdeen DA

Here are two situations that are pretty common when working with AI products. In the first, the MVP went well, the dashboard is green, and someone has to decide whether real users should see this. In the second, the thing is already in production, everyone agrees it could be better, and nobody can say which change would actually help.

Have you ever been in either of these? In this talk I make the case for a quality workflow in agent-based applications. This will be a field report from building agents with several customers, where I distill what has worked for me and what hasn't. So prepare for a very opinionated talk, where I put emphasis on the tools I use, the data governance practices I rely on, the distinction between technical and qualitative metrics, and how a good evaluation framework can give you direction when you most need it.

I hope this can also help you along the way in your current and upcoming projects. And if you are not a practitioner, it will give you a glimpse of the most important bit of developing AI products.