Attention founders who already tried an AI agent: read this before you build another one

Your AI agents didn't fail. You hired them without the right job description.

Agent Onboarding is an AI coach that maps your real week, picks the three jobs you should hand off first, and walks you through onboarding an agent to each one. A written Job Card and a short video at every step.

No prompts to learn. No code.

You did what everyone told you to do.

You set up an AI agent. Maybe for your inbox. Maybe for lead follow-up. You gave it access to everything and told it to be careful.

And for a while, it worked. It got you 90% of the way.

Then the last 10% showed up. An expired login. A webhook that failed overnight. Two agents that work off the same data and still don't agree.

So now you check every draft before it goes out. You wonder what it touched while you weren't looking. And you have no idea what last month's runs actually cost you.

You wanted an employee. You became its middle manager.

It's not the agent. It's the hire.

Think back to the last time you hired a person.

You didn't hand them the keys to everything and say "be careful." You told them what the job is. What they own. What they can touch, and what they can't. What "done" looks like. How much they can spend. Who they go to when something breaks.

Your agents got a prompt. Maybe a long one. But a prompt tells an agent how to do a task. It doesn't tell it what the job is.

That's the gap. Most tools start with the agent, not the job.

The 5-Line Job Card: five lines every new hire needs before day one

Every problem you have right now is one of these lines missing.

1

What it owns.

Nobody wrote this down, so your agents work off the same data and still don't stay in sync.

2

What it can touch.

Nobody wrote this down, so you're scared of what it'll do.

3

What "done" means.

Nobody wrote this down, so you review everything it makes.

4

A spending limit.

Nobody wrote this down, so you have no idea what it cost.

5

Who it calls when something breaks.

Nobody wrote this down, so the last 10% keeps landing on you.

A smarter agent won't fix that. A smarter agent without the right job description just makes the same mess faster.

So which AI employee should you hire first?

It's the question founders ask most. Inbox? Lead follow-up? Bookkeeping?

You can't answer it from a list of tools. You answer it the way you'd plan a first hire: by looking at your own week. What eats your hours. What only you can do. What someone else could own tomorrow, if the job were written down.

And you already know how to do that part. You don't need to learn prompts or code. You need to say what the job is. The chat does the rest.

Why the usual fixes keep failing

You've probably tried at least one of these:

None of these write the job down. That's the only fix that stays fixed.

Introducing Agent Onboarding

The heart of the AI Delegation Map. An AI chat that already knows how to onboard an agent, plus short videos that show you every setup step. You talk. It writes the job down. You follow the video. Job first. Agent second.

1

Your Delegation Map

Everything you do, sorted into what to hand off first, later, or never.

3

5-Line Job Cards

One written job for each of your three highest-value handoffs.

3

Onboarded agents

Set up one at a time, each with its own step-by-step video.

Example conversation

CoachWalk me through yesterday. What did you do, from the first email on?

Youok. answered customer emails for an hour, chased two unpaid invoices, client calls all afternoon. and i forgot three leads again

CoachGot it. Three jobs an agent could own tomorrow: inbox triage, about 5 hours a week. Lead follow-up, about 4 hours a week. Invoice chasing, about 3 hours a week. Which one annoys you most?

Youthe leads. i lose money there

CoachThen that's your first hire. Here is line 1 of its Job Card. Change any word you like.

  1. 1
    What it ownsThe first reply to every new inquiry, within one hour.
Example conversation. The coach works from your week, so your jobs and hours will differ.

Here's exactly how it works

Step 1: The chat maps your real week

  • Talk through everything you do, in your own words. No tool list, no prompts.
  • See every task sorted into what to hand off first, what to hand off later, and what stays yours.
  • Find the hours you didn't know were hand-offable.

Step 2: It picks your three first jobs

  • Answer "which AI employee should I hire first?" from your own week, not a generic list.
  • Choose the three jobs worth the most to you, and see why they beat the rest.
  • Leave the risky jobs for later, on purpose.

Your Delegation Map

Example

Your whole week, sorted by the coach. You see what to hand off first, and why.

Hand off first

  • Inbox triageSort, flag urgent, draft replies5 h / week
  • Lead follow-upFirst reply within the hour4 h / week
  • Invoice chasingRemind every unpaid invoice3 h / week

Hand off later

  • Weekly numbersNeeds clean data first2 h / week
  • Meeting notesLow value until calls are recorded1.5 h / week
  • Social repliesTone is still yours2 h / week

Stays yours

  • Sales callsTrust is the product6 h / week
  • Pricing decisionsOne wrong number costs real money1 h / week
  • HiringNobody hires your team for you1 h / week
Example map. Your tasks and hours will differ.

Step 3: It writes a 5-Line Job Card for each

  • What it owns, so your agents stop working off the same data and drifting apart.
  • What it can touch, so you stop wondering what it did while you weren't looking.
  • What "done" means, a spending limit, and who it calls when something breaks.

Step 4: You onboard each agent with a short video

  • Follow along, one agent at a time. Each video ends with the agent running on its card.
  • Stuck on a login or a failed webhook? Ask the chat. It has your cards in front of it.
  • Finish with a written job for every agent, so you can read, change or fire any of them.

Is this for you?

This is for you if

  • You run the business and you're the bottleneck in it.
  • You tried an AI agent (or three) and got 90% of the way.
  • You check every draft because you don't fully trust it yet.
  • You'd rather say what the job is than learn prompts.

This is not for you if

  • You want a prompt course or a tool tutorial.
  • You want an agent with full access and no written limits.
  • You want an agency to build and run everything for you.
Real example · we run this ourselves

What a job looks like when it's written down

We don't sell theory. This is our own inbox agent. It runs for us every day. Seven inboxes. One agent. One card.

5-LINE JOB CARD · INBOX AGENT real · runs every day
  1. 1
    What it ownsEvery new email in our 7 inboxes. It sorts each one: urgent, today, or ignore.
  2. 2
    What it can touchIt reads. It never sends an email. It never opens login codes.
  3. 3
    What "done" meansThree runs a day: 07:30, 13:30, 19:30. After each run, urgent mail is on Jan's phone within 5 minutes. Mail for today waits in one evening list.
  4. 4
    A spending limit$0 extra. It runs on our flat AI plan and has no tools to spend with.
  5. 5
    Who it calls when it breaksA second agent checks it every morning at 09:00. If the daily list is more than 26 hours old, it raises the alarm. Jan does not have to notice it himself.

Nov 2 to Nov 9, 2026 · eight days

5 urgent emails reached Jan's phone. 48 waited in the evening list.

First number: urgent, pushed to the phone. Second number: waited in the evening list.

  • Nov 20 · 5
  • Nov 30 · 7
  • Nov 40 · 12
  • Nov 52 · 6
  • Nov 62 · 7
  • Nov 70 · 3
  • Nov 80 · 2
  • Nov 91 · 6

One job, eight days, our own inboxes. Not a customer result.

Who's behind this

We're Jan and Martin. We don't teach AI agents from a whiteboard. We run our own companies on them.

"I spent sixteen days with a fever, some of it in a hospital bed. Every morning I opened Stripe on my phone. Sales were still coming in. Our agents processed them, wrote the reports and answered the support requests."

Jan

That's what we want for you: agents doing their job, because someone finally wrote it down.

Plus three bonuses included at $47

+

The Blank Job Card Pack

Fill-in 5-Line Job Cards. Write the next agent's job yourself in minutes.

+

The Agent Spend Tracker

A simple sheet that shows what each agent cost you each month. No more "I have no idea."

+

The Monthly Agent Review

A one-page checklist to keep, retrain or fire each agent, like a real performance review.

Get Agent Onboarding

  • The AI coach: maps your week and picks your three first jobs
  • Three 5-Line Job Cards, written with you
  • Step-by-step videos to onboard each agent
  • Bonus: Blank Job Card Pack
  • Bonus: The Agent Spend Tracker
  • Bonus: The Monthly Agent Review
  • You need: your own Claude plan. Pro is enough.

One payment to us · no subscription to us $47

Get Agent Onboarding · $47 One payment. Secure checkout.

30-day money-back guarantee · No prompts to learn · No code

The 30-Day "Doing Its Job" Guarantee

Try Agent Onboarding for 30 days. If your first agent isn't doing its job, or you decide it's not for you, email us and you get every cent back. You don't have to prove anything or fill in a form.

The Job Cards you wrote are yours to keep.

Questions

I'm not technical. Can I still do this?

Yes. You don't learn prompts and you don't write code. You tell the chat how your week runs and say what each job is. The videos show every setup step on screen.

Which tools do I need?

A Claude subscription. The coach runs inside Claude, and so do the agents in our setup videos. The Pro plan is enough. It costs about $20 a month, paid to Anthropic, not to us. Beyond that you only need the tools the job already uses, like your inbox or your calendar. No code.

What does it cost to run my agents?

They run inside your Claude plan. There is no bill per run. And line 4 of every Job Card is a spending limit. So next month's cost is a number you wrote down, not a surprise.

I already use ChatGPT, n8n or Zapier. Does this still work?

The Job Card does. It is plain text, and it works for any agent on any tool. The coach and the setup videos run in Claude, because that is what we run ourselves. So you need Claude for those two parts.

How is a job description different from a prompt?

A prompt tells an agent how to do a task. A job tells it what it owns, what it can touch, what "done" means, how much it can spend, and who it calls when something breaks. Those five lines are the 5-Line Job Card.

I already built agents that didn't work. Do I start over?

You start with the job, not the agent. The chat maps your week, picks the three jobs worth the most, and writes the card for each one. Agents you already have are judged against the card like any new hire.

How does the AI coach decide which three jobs to hand off?

From your own week. It looks at what eats your hours, what only you can do, and what someone else could own tomorrow if the job were written down. Then it picks the three worth the most to you.

What if something breaks while I set up?

Ask the chat. It has your Job Cards in front of it. And line five of every card names who gets called, so a failure doesn't default to you.

How long does it take?

You set the pace. First the coach maps your week and writes your three Job Cards with you. Then you set up one agent at a time, each with its own short video. There is no schedule to keep up with.

What if it's not for me?

Try it for 30 days. Not for you? Email us and you get a full refund of the $47. No reason needed.

Stop being its middle manager. Write the job down.

Get Agent Onboarding · $47 30-day money-back guarantee

P.S. Your agents didn't fail. They were never told what the job was. Write the job down first. Build the agent second. Start with your first three jobs.