Meet Notion's Custom Agents. They Never Sleep.

Notion has opened “Custom Agents” to everyone, pushing agents from a novelty into a daily workplace layer that can answer questions, update knowledge, and reduce repetitive coordination work. A flagship example is “Smilers,” an agent designed for Notion’s internal environment team: it shows up first in Slack when employees ask questions, pulling answers from an “environment home” database and a dedicated knowledge database built for Smilers. When employees provide feedback in Slack, Smilers can feed that information back into the knowledge base in Notion—removing a chunk of manual upkeep from the team while keeping responses consistent across offices.

The environment team also plans to use Smilers for pattern detection and proactive operations. One near-term workflow is trend forecasting: Smilers will flag recurring feedback themes when the same issue shows up three to five times, helping the team spot systemic problems rather than reacting one-off. Another step is initial outreach—drafting emails to building management when recurring concerns surface—so the team can move from “someone asked” to “we’re acting” without waiting for extra cycles of triage.

Beyond Smilers, the transcript highlights how custom agents are being used as personal and team “command centers.” One builder describes running roughly 20 agents, mixing personal and team versions, each with themed “helmets” tied to their roles. The workflow starts with capturing tasks and project context throughout the week, then using agents to automate daily updates and generate weekly summaries. In practice, this means the system can compile what happened, what tasks were closed, what projects advanced, and what remains open—turning scattered inputs into structured planning and reporting.

Several builders emphasize that custom agents lower the barrier to automation. One non-engineer says setup is largely driven by talking to Notion AI with the right context, letting integrations and flows get configured without coding. Engineers echo the same theme: custom agents reduce the time cost of building and maintaining automation that would otherwise fall to traditional engineering work.

A concrete engineering use case comes from a “test triager” agent. Historically, bugs were handled by reacting in Slack and filing tasks via custom code. The agent upgrades that process by checking the knowledge base and task tracker first: if a reported bug already exists, it responds to the reporter with status details (for example, that the bug is assigned to a specific person and already being worked). If it’s new, it files the appropriate task. The payoff is less duplicate work and fewer “ignored” reports—because acknowledgments happen quickly via reaction triggers, letting teams keep moving without constant context switching.

Overall, Custom Agents are positioned as always-on assistants that combine knowledge retrieval, workflow automation, and feedback loops—aimed at making internal support, planning, and triage faster and more equitable across locations.