The number one missing piece to AI employees that actually work (and it isn't another skill)

The Incuv8or AI Framework - step 5 of 6: Hand out the work from your task manager. Before this: add employees only when they pass the test. Next: add routines where they're needed.

So many people online are talking about AI agents. How to write a skill, how to connect it to Claude, how to set up a routine. I've written those posts too. But almost no one is talking about creating a workflow, and I think that's the number one missing piece to AI agents that actually work.

When OpenClaw came out, we were all excited that we could finally get AI to do stuff for us. You'd give it a brain dump of everything you wanted done, and because it had no task management system, it would get lost and stop. I had the same thing happen in Claude. I'd have a great employee, with a great skill, and it would sit there doing nothing until I opened a chat and told it what to do.

When I set up the workflow, it made everything else make sense. Now I brain dump into a task in ClickUp, pick the employee, set the status, and the work gets done in the background while I get on with my day. It's back in my inbox when it's finished, with the whole story on the task.

This post is the strategy behind that, so you understand why it works, and then the setup, with my ClickUp template so you don't have to build it from scratch.

Why your AI employees aren't doing anything

An AI employee knows how to do its job. Its skills tell it how to write your blog post, sort your inbox or draft a proposal. What it doesn't know is what to work on today.

Every time Claude starts, it starts with no memory. It doesn't know what you asked for yesterday, what it finished last time, or what you sent back with notes. If the only place that information lives is a chat window, it's gone the moment you close it.

A skill tells an AI employee how to do the job. The workflow tells it which job, whose turn it is, and when to stop and hand it back to you.

That's why a pile of agents doesn't do anything on its own. They're like a team of very capable people who turn up every morning with no task list, no manager and no idea what happened yesterday.

The strategy, in four parts

This is the part I wish someone had explained to me at the start. Once you see these four pieces, every workflow tool makes sense, and you'll understand why each part of the template is there.

1. Every piece of work is a task

The task is the brief, the memory and the record, all in one place. The description or the activity feed holds what you want. Every comment stays on it, so when an employee picks it up again, it can read everything that's happened so far. When it comes back to you, you can see what was done without asking.

This is what makes the work survive. A chat is gone by morning. A task is still there.

2. The status says whose turn it is

You don't need a status for every stage of work. You need to know, at a glance, whether something is with you or with an AI employee.

That's why my statuses start with "w/". W/ agent to do means it's the AI's turn. W/ me for review means it's mine. When an employee finishes, it moves the task to my side, and when I've added my notes, I move it back to theirs with w/ agent for edits. Every handover is one status change.

The statuses are also your guardrails. Nothing gets published or sent unless I've moved it to w/ agent to publish, so the AI never makes that decision for me.

3. A field says who does it

One dropdown field, called AI EMPLOYEE, with one option for each employee on your org chart. That's how you give a task to an AI without paying for a seat for it. Any task, on any list, can go to any employee by picking a name.

The field says who. The status says when. Between them, an employee always knows exactly which tasks are its own.

4. Something has to check for work

This is the piece most people never build. Your task manager holds the work, but something has to go and look for it.

Mine is a small Python script called the dispatcher. Every 10 minutes it checks ClickUp for any task with an AI EMPLOYEE set and a status that starts with "w/ agent", and starts Claude on it with that employee's agent file. When the employee's finished, it puts the work on the task and moves it to my side.

That's the loop. I don't open a chat, I don't remember to run anything, and it works while I'm out.

Why I use ClickUp

I use ClickUp because I was already running my business on it. My tasks, my content, my client portals, my projects and my CRM are all in there. I didn't want one system for my tasks, another for my clients and another for my AI employees. I want it all in one spot, and that's how my brain works.

It also means I can add my clients and human team members to the same place as my AI employees, and it's a reputable platform that already has the security it needs. I'm not a developer, so I'd rather use that than a system I've vibe coded myself.

If you use a different task manager, it can do the same job as long as it has statuses, custom fields and comments. The steps below say ClickUp.

What you need before you start

• A ClickUp account.
• The Claude desktop app on a paid plan, with the Claude Code tab.
• Your AI brain, so your employees know your business: how to set up your AI brain in Obsidian.
• At least one AI employee. If you haven't got one, build a COO first: the first AI employee I'd build.
• My ClickUp template: Download the Incuv8or ClickUp template →

You don't need to know how to code. Claude writes the one small script this needs.

Step 1 - Connect Claude to ClickUp

Open the Claude desktop app and connect ClickUp.

1. Click Settings, then Connectors.
2. Find ClickUp and click Connect.
3. Sign in on ClickUp's own screen and allow access.

You sign in on ClickUp's screen, so there's no password to paste anywhere. Then click the Claude Code tab, start a new chat, click the + at the bottom and choose your AI brain folder.

Step 2 - Add the template to your ClickUp

I want to add the Incuv8or ClickUp template to my workspace. Here's the
link: [PASTE THE TEMPLATE LINK]. Walk me through it one step at a time &
tell me exactly where to click.

When it's in, you'll see a list called 🤖 AI Workforce with the statuses already set up, and a 👋 Start here task that explains every part of it. Open that task first.

Step 3 - Map one job onto the workflow

Before you hand anything over, pick one job you do every week and walk it through the statuses. This is how you'll see where the AI does the work and where you need to check it.

Read my AI brain & my org chart. I want to hand one job to my AI
employees through ClickUp: [THE JOB, e.g. writing my weekly newsletter].

Walk the job through my statuses (to do, w/ agent to do, w/ me for
review, w/ agent for edits, w/ agent to publish, done) & tell me which
employee does each part, where I need to check it & what the employee
needs in the task description to do it without asking me. Then write me
a task brief template for this job that I can copy into ClickUp.

What you'll see: the job broken into turns, yours and the AI's, and a brief template for the task. Save the template in your AI brain, because you'll use it every time.

Step 4 - Build the dispatcher

This is the script that checks ClickUp for work.

I want my AI employees to pick up work from ClickUp on their own. Write
me a small Python script, the dispatcher, that runs every 10 minutes on
this Mac.

Every run, it checks ClickUp for tasks where AI EMPLOYEE is set & the
status starts with "w/ agent". It works on no more than 2 tasks per run,
one at a time. For each one, it starts Claude with that employee's agent
file & the task, including every comment on it. When Claude's done, it
puts the work on the task & sets the status to "w/ me for review".

My ClickUp key goes in my Mac's Keychain, never in the script & never in
this chat. When it's time, tell me where to create the key & open the
place it goes for me. Explain each step in plain English & do any setup
yourself.

A key is a password for your ClickUp account that a script can use. The Keychain is the password store that's already built into your Mac.

Claude tells you when to do each of these.

1. Claude sends you to the page in ClickUp's settings where you create a personal API token. Create one and copy it.
2. Claude opens the box on your Mac where the key goes. Paste it there, on your own screen.
3. Go back to the chat and tell Claude it's done.

If a key ever ends up in a chat, delete it in ClickUp and make a new one.

Step 5 - Give it one task

Open the example task in the template. Brain dump what you want into the description or the activity feed, the way you'd brief a person. Set AI EMPLOYEE to your COO and the status to w/ agent to do.

Within 10 minutes the dispatcher picks it up. When it's finished, the task is at w/ me for review with the work on it. If you want changes, comment on the task and set it to w/ agent for edits. If you're happy, close it.

The catch, and the fix

When I first started doing this, if I had fifteen tasks allocated to the agent, it would try to do all fifteen at once. It blew my rate limits and everything crashed.

The fix is the line in the Step 4 prompt that says no more than 2 tasks per run, one at a time. I can allocate as many as I like, and they get worked through without hitting my limits.

The second catch is that the dispatcher runs on your computer. If your laptop is shut, nothing gets picked up until you open it again. Mine now runs on a Mac Mini that stays on all the time, so the work keeps moving whether my laptop is open or not.

What this looks like as a business

Most business owners being sold AI agents have the same problem I had: great agents that don't do anything without them. Setting up the workflow, in the task manager they already use, is the piece that makes the agents work. You're paid for knowing what to build and in what order, and this is the piece most people skip.

Start with one

This week, add the template, read the Start here task, and walk one job through the statuses with the Step 3 prompt. When the brief template for that job looks right, build the dispatcher and hand over your first task.

If you haven't built your first employee yet, that's the step before this one: build an AI COO first.

If you haven't got your AI brain set up, that's the foundation for all of this: start here.