On September 17, we hosted Exploring Discourse Workflows, a practical look at how communities can use Discourse Workflows to automate multi-step processes directly in Discourse.
Watch the replay:
Workflows are visual automations that respond to something happening in your community and then take action.
At their simplest, you can think of them as trigger → action → result.
More complex workflows can add conditions, branching, stored data, AI, scheduled actions, and connections to external services.
During the session, we broke down the basic building blocks of a Workflow, including:
- Triggers: What starts the Workflow, such as a new topic, a scheduled event, or an incoming webhook
- Conditions: Rules that determine which path the Workflow should take
- Actions: What Discourse should do next, such as creating a post, granting a badge, notifying someone, or calling an external API
- Flow-control and utilities: Additional building blocks that help connect and manage more complex automations
We also explored a few capabilities that become especially useful when combined: AI categorization, data tables, and group automation. These can help communities classify content, store information a Workflow will need later, and route or manage users based on defined conditions.
Workflows in action
A big part of the session focused on moving beyond the building blocks and looking at actual community problems that could be automated.
A few examples included:
- Automated reporting and follow-up: Run recurring Data Explorer queries, summarize the results, and create reports for different teams.
- Getting the right question to the right person: Identify what a new topic is about and route or escalate it to the person or group best equipped to respond.
- Closing the loop on unanswered discussions: Find topics that still need attention, remind people to mark solutions, or escalate conversations that remain unresolved.
We also demonstrated how several pieces can work together in a single automation. For example, a Workflow can use information stored in a data table alongside AI to determine which subject-matter expert should receive a question. Scheduled Workflows can also run Data Explorer queries, summarize the results, and send recurring reports into a topic or chat.
Start with the problem, not the nodes
One of the biggest takeaways from the session was that you don’t need to start by understanding every available Workflow node.
Instead, start with the thing you want to improve.
Ask:
What starts it?
Is it a new topic? A schedule? A user action? Something coming from another system?
What should Discourse check?
Content? A user or group? A query result? Information saved earlier in the Workflow?
What should happen next?
Should something be tagged, sent, updated, routed, delayed, reminded, or escalated?
If you’re not sure what you would automate yet, we shared four places to look for ideas:
- The repeat: What are you or your team doing over and over?
- The friction: What keeps not happening the way you want?
- The steal: What has another community done that made you think, “Oh, that’s smart?”
- The weird: Could automation create a useful reminder, unlock, consequence, game, or even an inside joke?
Not every community process can be automated with Workflows today, and we also discussed some current limitations and ideas for future additions. But the goal isn’t necessarily to automate everything. It’s to identify the repetitive, easy-to-miss, or time-consuming processes where automation could make running your community a little easier.
cc 
