How is your team organized around Dust ? ⤵️

👋 Hi everyone, curious how your teams are organizing around Dust. At Alma, we’ve taken a fairly open approach: giving people broad access to build both agents and skills. This has encouraged creativity, but also led to a growing number of agents with uneven AI maturity across teams. We have achieved very strong adoption, but the complexity of AI agents remains relatively low (e.g. limited webhooks, agents that call agents, etc). Training helps, but protected time to build and maintain is also proving important. I’m curious about the alternatives: do you have embedded builders build protected build time, a centralized team of “power builders” supporting other teams, or a more organic model? How do people bring forward use cases, and what have you learned about balancing autonomy, quality, and adoption?

Hi Alexis ! Adoption is indeed a great first step. The harder question is: what comes next? In our experience with our clients on Dust, the answer lies in combining clear governance with dedicated time to support employees over the long term. On the governance side, it is important to set up a clearly identified project team with distinct roles:

Builders represent their teams, identify relevant use cases, train their peers, and make sure agents are properly shared and adopted within the team. They also work with team managers to review and approve the agents that are ready to move into production. For this model to work, builders need dedicated time in their role to build, iterate, and maintain agents. This work should also be recognized and rewarded as part of their responsibilities. Another key role is the AI Ops technical lead. What we have found to work particularly well is continuous support through:

From a governance perspective, the project manager should also maintain a clear overview of the agents being used and adopt a product-oriented approach to their lifecycle:

It is also important to track the ROI of agents in production, taking into account their credit cost, usage, time saved, and day-to-day impact on the team. The next challenge is moving beyond isolated individual use cases and rethinking workflows at the team and company level. This is where AI adoption can evolve from experimentation into a sustainable operating model