Automate Workflows with Event-Driven AI Agents

Learn how to create AI agents that watch for triggers, connect to integrations, and take actions automatically — from meeting prep to lead qualification.

AI Agents are automated workflows that monitor for specific events and take action when those events occur. Unlike AI Assistants — which respond when a person asks a question — AI Agents run in the background, watching your connected tools and acting on your behalf. When a trigger fires, an agent reads the relevant context, processes it with AI, and executes one or more actions: drafting a message, updating a record, sending a notification, creating a task, or generating a report.

Agents are built for workflows that are repetitive, time-sensitive, or depend on staying current with activity across multiple tools.

AI Agents vs. AI Assistants

	AI Assistants	AI Agents
Triggered by	A human asking a question	An event or schedule
Runs	On demand	Automatically in the background
Output	A response in a chat interface	Actions across connected tools
Best for	Q&A, knowledge retrieval	Automation, monitoring, follow-through

Creating an agent

1. Open the Agents section

   Click Agents in the left sidebar, then click New Agent.

2. Name and describe the agent

   Give the agent a descriptive name — for example, “Meeting Prep Agent” or “Lead Qualification Agent”. Add a short description of what the agent does so your team understands its purpose.

3. Choose a trigger

   Select the event that starts the agent. Triggers include:

   • Calendar event created or updated — fires when a meeting is added or changed
   • Email received — fires when a matching email arrives in a connected Gmail or Outlook account
   • CRM record updated — fires when a lead, deal, or contact changes in a connected CRM
   • Slack message posted — fires when a message appears in a specified channel
   • Jira issue created or updated — fires on ticket activity in a connected Jira project
   • Scheduled time — fires at a set time each day, week, or custom interval

   Configure the trigger’s filters to narrow when it fires — for example, only calendar events with external attendees, or only CRM leads with a score above a threshold.

4. Define the agent's actions

   Add one or more actions the agent takes when triggered. Available actions depend on your connected integrations:

   • Draft and send a message — Slack DM, email, or team message
   • Create or update a record — CRM contact, Jira issue, Notion page
   • Generate a document or report — in WorkLLM or a connected tool
   • Post a summary — to a Slack channel, email thread, or shared WorkLLM thread
   • Add a calendar event note — attach AI-generated prep notes to a meeting
5. Connect your integrations

   If the trigger or actions require an integration (Gmail, Google Calendar, Slack, Jira, a CRM, etc.), connect them in Settings → Integrations before creating the agent. Once connected, they appear as options in the trigger and action pickers.

6. Configure the AI prompt

   Write the instruction that tells the agent what to do with the context it gathers. This works like a system prompt — describe the task, the expected output, and any constraints.

   Be specific about the output format. If the agent is generating a briefing document, specify the sections you want. If it is qualifying leads, specify the criteria it should evaluate.

7. Test and activate

   Click Test Agent to run a simulation using sample data from your connected integrations. Review the output and adjust the prompt or actions as needed. When you are satisfied, click Activate.

Example agents

Meeting prep agent

Trigger: Calendar event created with external attendees

What it does: When a new meeting is added to your calendar with external guests, the agent pulls context from your CRM (company info, recent activity, deal stage), generates a briefing document, and attaches it to the calendar event — ready before the meeting starts.

Actions: Fetch CRM records → Generate briefing → Attach to calendar event → Post summary to Slack

Lead qualification agent

Trigger: New CRM lead created

What it does: When a new lead is created in your CRM, the agent evaluates it against your qualification criteria, generates a score and summary, updates the CRM record, and routes high-priority leads to the sales Slack channel for immediate follow-up.

Actions: Evaluate lead data → Score and summarize → Update CRM record → Post to Slack if high-priority

Daily digest agent

Trigger: Scheduled — every weekday at 8:00 AM

What it does: Each morning, the agent pulls activity from the past 24 hours across Slack, Jira, and your CRM, summarizes the key updates, and sends a digest to each team member’s email or Slack DM.

Actions: Fetch activity logs → Summarize by category → Send personalized digest via email or Slack

Customer follow-up agent

Trigger: Email received matching subject or sender criteria

What it does: When a customer email arrives that matches a pattern (subject contains “renewal”, sender is a key account), the agent drafts a personalized follow-up response using CRM context and sends it to the account owner in Slack for one-click approval before sending.

Actions: Parse email → Fetch CRM context → Draft response → Send to account owner for approval

Monitoring agent runs

All agent activity is logged in the Agent Runs tab. For each run you can see:

• The trigger event that fired
• The timestamp and duration
• The input context the agent processed
• The actions taken and their results
• Success or failure status, with error details if applicable

Use this log to verify your agents are running as expected and to debug issues.

Managing agents

Edit an agent

Click the agent name in the Agents list to open the editor. Adjust the trigger, prompt, or actions and save changes. Updates take effect on the next run.

Pause an agent

Toggle the Active switch on any agent to pause it. The agent retains its configuration but stops running until you re-enable it.

Delete an agent

Open the agent editor and click Delete. Deleting an agent stops all future runs and removes its run history. This action cannot be undone.

Only the agent creator can edit, pause, or delete an agent. Agents that connect to shared integrations (such as a team Slack channel or shared Gmail inbox) may affect teammates. Review the scope of an agent’s actions before activating it.