Slack wants to be where workflows and AI agents finally meet. On 18 November 2026 the company added agent steps to Workflow Builder, event-triggered Slackbot Tasks and scheduled Agent Tasks, so automations can hand work to Slackbot or any third-party agent and bring the result back, with a human review step built in before anything reaches tools like Jira, Google Sheets or Salesforce.
Quick Take
- Agent steps in workflows: new “Start conversation with Slackbot” and “Start conversation with Agent” steps hand work to Slackbot or any third-party or custom agent.
- Slackbot Tasks: start automatically when something happens in Slack, such as a new message, a reaction or an updated list item.
- Agent Tasks: save a prompt to an agent and run it on a schedule or on a Slack event.
- Human review: Workflow Builder automatically adds an editable review step after every AI step.
- Availability: Workflow Builder is included in Pro, Business+ and Enterprise plans; no code is required.
In plain English
Until now, teams often used automations for routine steps and AI agents for writing or research, then copied text between them by hand. Slack is joining the two. A workflow can now ask an agent to draft something, wait for a person to check it, and then pass it on. You can also tell Slackbot to do a job every time a certain thing happens, or have an agent run the same prompt every week.
What happened
“Today, we’re making Slack the automation hub for the agentic era,” the Slack team wrote. “Workflow Builder, Slackbot, and your AI agents now work together in one place, so the thinking happens inside the process, not beside it.”
Slack says agents in workflows already know the team’s context and respect its permissions. For developers, the Slack changelog adds a Tasks tab in the DM of agents that use the Agent messaging experience. It is on by default, and users can hand an agent recurring work like a morning digest or weekly report.
How it works
- In a workflow: add the step, write the prompt once, and add workflow variables, skills and knowledge sources. When the workflow reaches the step, Slackbot opens in the person’s sidebar with the prompt already sent.
- Example from Slack: an IT help desk form posts troubleshooting steps, and a “Need more help” button pulls in a Help Desk agent with the question already sent.
- Slackbot Tasks: describe the task to Slackbot. Calendar, Salesforce and service triggers from tools like PagerDuty are coming next, Slack says.
- Agent Tasks: open the agent’s DM, add a task, write the prompt and pick when it runs. Results land in a thread you can continue.
- For app builders: the Tasks tab can be turned off in app settings or by setting features.app_home.agent_tasks_enabled to false in the manifest; it requires the agent_view feature and the assistant:write scope.
Why it matters
For teams, recurring jobs such as stand-up drafts, status reports or ticket summaries can run without someone remembering to start them, while a person still approves what goes out.
For developers of Slack agents, scheduled tasks arrive without extra work: users can put an existing agent on a schedule from its DM.
For the market, Slack is positioning itself as the coordination layer for agents from many vendors, competing with similar moves by Netlify and others to put agents and business users in one place.
What’s still open
- How AI steps and agent runs are priced or limited by plan.
- When the announced calendar, Salesforce and PagerDuty triggers arrive.
- How admins audit what agents do across many automated tasks.
Sources
- Slack, “Bring Agents Into Your Workflows, Right in Slack,” blog post, 18 November 2026
- Slack developer changelog, “Introducing the Tasks tab for agents,” 18 November 2026
i10x Perspective
The automatic review step is the smartest part of this launch. Most companies are not ready to let agents write straight into Jira or Salesforce, and a built-in pause makes the first rollouts much easier to approve. More agent tools in our Agents section.



