# AI-native workflows

## Automate the impossible. Grow 100x and beyond.

Map a job once on one canvas. Brain² and Super Agents run the steps, you approve the result, and every run builds on the last.

## Beta in ClickUp Labs

### Weekly status digest

- **Steps**
- **People**
- **Outputs**

**Website relaunch** · Ready  
**Run Workflow**

**Manual run**  
**Run Workflow**  
**Manual start**

### Condition
- Anything at risk? **Yes**/**No**

### Agent
- Ask owners for status **Super Agent**

### Wait
- Wait 1 day

## Owners reply

### MergePaths rejoin

### Agent
- Draft digest

### Brain²

### Approval
- Waiting
- Approved
- Changes
- Escalated  
- Review digest in a task

### Output
- Reviewed digest
- Ready to use

## Why AI-native

### Built to think, not just trigger

Rule-based tools pass data from app to app. AI-native workflows read the work, make the calls, and hand the result to a person.

### Rule-based automation
1. If this
2. Then that

- Moves data between apps
- Runs only the rule you wrote
- Breaks when the input changes

### AI-native workflows
1. Reads the work
2. Decides the path
3. Drafts the output
4. You approve

- Reads tasks, Docs and replies
- Picks the path when a run has to choose
- Learns from every run it finishes

## The canvas

### Give AI a repeatable set of tasks

Set up the steps once. Run the workflow again whenever you need it, and check what happened each time.

### Choose how it starts
- Run it yourself on any task, or set it to start on its own.
  - Manual run
  - Automation
  - Schedule
  - Webhook

### Hand the thinking to AI

Brain² or a Super Agent reads the work and writes the draft. When the run has to choose, AI picks the path.

### You make the call

Add an approval anywhere a person should check the work. They decide right in a ClickUp task.

### Keep the run moving

Give people time to reply and bring split paths back together. When it ends, you see exactly what the run made.

### Some steps don't need AI

When a step should happen the exact same way every run, use one of these instead.
- Transform
- Action

### Manual runRun Workflow

**On Website relaunch**
**Condition**  
Anything at risk? **Yes**/**No**

**Agent**  
Ask owners for status **Super Agent**  
**Wait**  
Wait 1 day

### Merge
- Paths rejoin

### Agent
- Draft digest

### Brain²
a

### Approval

- Waiting
- Approved
- Changes
- Escalated  
In a task
- Approved
- Changes

### Output
- Reviewed digest
- Ready to use

## Transform
- Keep the summary
- Top 3 lines

## Action
- Add a comment **On Website relaunch**

1. Choose how it starts. Run it yourself on any task, or set it to start on its own. Steps: Triggers.
2. Hand the thinking to AI. Brain² or a Super Agent reads the work and writes the draft. When the run has to choose, AI picks the path. Steps: Agent, Condition.
3. You make the call. Add an approval anywhere a person should check the work. They decide right in a ClickUp task. Steps: Approval.
4. Keep the run moving. Give people time to reply and bring split paths back together. When it ends, you see exactly what the run made. Steps: Wait, Merge, Output.
5. Some steps don't need AI. When a step should happen the exact same way every run, use one of these instead. Steps: Action, Transform.

## Learning loops

### Every run makes the next one better

Each run saves what worked. The next run starts from those learnings, so the work compounds instead of starting over.

### Learnings applied
1. Run 3 Carousels outrun single images
2. Run 2 Tuesday 9 AM beats Friday afternoon
3. Run 1 Customer quotes get 2x the saves

#### Run
43 learnings saved

### Agent
- Draft posts

### Brain²

### Approval
- Approve In a task

### Publish
- Posts go live
- LinkedIn, X

### Agent
- Measure results
- Super Agent

### Learn
- Save what worked
- To a Doc

A loop: Brain² drafts posts, a person approves them in a task, the posts go live, a Super Agent measures the results, and what worked is saved to a Doc. The next run drafts from those learnings.

## Human in the loop

### AI does the work. You make the call.

Add an approval anywhere a person should check the work. The run waits in a ClickUp task until someone decides.

- **Approve** The run moves on
- **Request changes** AI revises with your note
- **Escalate** Someone else decides

### Task
- **Review LinkedIn posts**  
Waiting on Sam

Brain² drafted 3 posts in run 4

> We cut our launch checklist from 40 steps to 6. Here’s the workflow that does the rest.

## For every team

### Every team has a job worth repeating

Pick a team to see a workflow in action, from the starting task through AI steps to a person’s review.

**Marketing**  
**Product**  
**Sales**  
**Support**

### Launch kit

**Native Sheets launch** · Ready

### Manual run
- Run Workflow
- On the launch task

### Agent
- Collect launch details **Super Agent**

### Agent
- Draft the blog post

### Brain²
a

### Agent
- Draft the email

### Brain²
a

### Agent
- Draft social posts

### Brain²

### Merge
- All drafts in
- Waits for all 3

### Approval
- Waiting
- Approved
- Changes
- Escalated  
- PMM review
- In a task

### Output
- Launch kit
- 3 assets, ready to use

## Integrations

### Bring in the other apps you use

Connect an app once in App Center, then choose it inside any Agent step. Apps with an MCP server connect the same way.

- Salesforce
- ZoomInfo
- Attio
- DocuSign
- Box
- Webflow
- PostHog
- Tableau Cloud
- Ttl;dv
- Calendly
- GitHub
- Any MCP server

## AI transformation

### Want us to set it all up?

Our team maps your jobs, builds the workflows and Super Agents, and gets your people running. You skip the setup.

### AI Workflows

## Set it up once, run it with a click

Try AI Workflows

### Beta in ClickUp Labs

### Manual run
- Run Workflow
- Pick an input task

### Output
- Reviewed digest
- Approved by Sam

### Made with love in ClickUp
**Made within ClickUp**
