This is the part most people are never shown
The install is done. Foundation is in. The agents have read your data. Now what do you actually say to them on a Tuesday morning? That is this module, and I am going to show you with one business so it is not abstract.
The business is Clear Launch Co
It is a made up course business, built so I could break things without hurting a client. Maya Reyes teaches people how to launch their first online course. She sells a $47 workbook, a $997 self paced course called Launch in 30, and a $4,500 six week group called the Launch Room. About $31,000 a month, a list of 6,200 on ConvertKit, one YouTube video a week, $4,800 a month in Meta ads to the workbook. One owner, a part time VA, a video editor and a bookkeeper. Pretty normal. Every scenario below is Maya using one agent on one real job. Eight of them are what the agents actually did in the demo Project, and the screenshots in this course come from those sessions. The other five, Funnel, Website, Proof, Finance and Strategy, are written from how those agents work, because the demo did not run them. Your business is different. The pattern is the same.
The pattern is this, for every agent
You open its session, or a new one if the old one has gotten long. You give it the first five percent: the idea, the specifics of this one job, and anything that changed since last time. It does the job using your profile, your plugged in tools and its own method. It comes back with the work and a short list of what it needs from you. You give the last five percent, which is a yes, a correction, or a go ahead. If you find yourself doing the middle by hand, something is missing from what it knows about you, and you fix that instead of doing the work yourself.
Three habits make this work
Talk to it like a capable new hire, with specifics, not like a search box. When you correct something, say why, because the why is what gets folded back into the rules. And when a job is done, close the loop: tell it done, or tell it what to change next time.
Read the scenarios first
Then the directory below, which has how to open each agent and three prompts to start with. Keep this page open for your first two weeks. Module 12 shows how the agents hand work to each other.
What you do, step by step
Pick the agent on the top line of your map.
Open its session. Run starter prompt one from its card below, with your real details filled in. Do the whole loop: read the output, correct one thing with a reason, approve the next step.
The Offer Agent finishing. What was agreed, what was not, what is still open, and one thing for Maya to rule on. The spec is saved to Project files on the right. 
The Email Agent’s Tuesday email, built from the video transcript. Nothing loads until Maya says so. The 1,273 credits on the right include the agent setting itself up on this first run. Run starter prompt two in the same session only if it is the same job.
If it is a different job, start a new session in the Project and say the agent’s name first.
Do this with each core agent on your map once in the first week.
One real job each, not a demo. Something you would have done yourself this week.
The Sales Agent sorting five DMs. Each one gets the offer that fits, the reply in Maya’s voice, why, and one thing to check before sending. 
The Operations Agent’s Monday briefing. Three things need Maya, with the money thread left undrafted on purpose. Then what the VA can take, then what can wait. 
The Delivery Agent’s welcome email for a new student, written for the VA to send. The Q&A date in this draft was a week off. Maya said which Thursday, it fixed the line and saved all three drafts. Each time you correct an agent twice the same way, say: turn that correction into a rule for next time.
It updates its own notes in the Project. This is how the agents get better without you re explaining.
At the end of the week, open the Project and read the session names.
If any session is a grab bag of three jobs, that is where your costs are hiding. Next week, one job per session.
One Business, Every Agent At Work
| Agent | The situation at Clear Launch Co. | What Maya says | What comes back, and what Maya does |
|---|---|---|---|
| Foundation | Foundation’s first pass called Maya a solo operator. True on paper, no employees. But it left out the VA, the video editor and the bookkeeper, and every agent would have written as if she did everything herself. | In the Foundation session: solo means no employees. The VA, the editor and the bookkeeper are real. Add them as a team note. | Foundation writes the team note into the profile and says which agents read it. Three sessions later the Delivery Agent drafts a welcome email for the VA to send, not for Maya, because it read the note. Nobody told it twice. |
| Offer Agent | Launch in 30 sells about 6 a month between launches. Maya wonders if she needs a new offer between the $47 workbook and the $997 course. | offer agent. Review my ladder against my map and Foundation. Do I need something between the workbook and the course? | A spec that says no. Three offers, same prices, and the job is to take evergreen from 6 a month to 20. It writes one rule every agent now follows: the course is open all year, so a deadline names the group start, never the course closing. Maya agrees. Saved to offer/offer-spec.md. About 700 credits. |
| Funnel Agent | The workbook to course sequence is five emails written in 2024. It converts about 9% and nobody has looked at it since. | funnel agent. Review the workbook to Launch in 30 path from the offer spec and the map. Pages, emails, where people leave. Nothing built yet. | A one page plan: the five emails rewritten around the deadline rule, one change to the workbook thank you page, a credit ceiling to build it. Maya says build it with a 2,000 ceiling. The agent hands the emails to the Email Agent and the page to the Website Agent. |
| Email Agent | Tuesday’s video is scheduled. The list has not heard from Maya in ten days, again. | email agent. This week’s email from Tuesday’s video transcript, in my voice. Show me before anything loads. | A 309 word draft built from the transcript, one link, no deadline line. It checked the link and found it broken, so it marked the email do not send. Maya approves the copy as written. Nothing goes into ConvertKit until the link is fixed. About 1,300 credits, because the first run also set the agent up. Later runs are a fraction of that. |
| Content Agent | Same Tuesday video. One video a week goes out and nothing else gets made from it. | content agent. Plan next week’s content from the transcript. A YouTube outline, three Instagram captions, one email idea. | A plan with every line traced to the transcript or to Foundation, quoting Maya’s own words. Three captions handed to the Image Agent. Maya approves the plan. Nothing is posted until she says so. About 370 credits. |
| Image Agent | The Email Agent found one line in the video that would make a good card. | To the Email Agent: hand this to the image agent. Then, in a new session: image agent. | It opens by saying it found the handoff. One 1080 by 1080 quote card in Maya’s colours: “The smaller the test, the faster you learn.” Saved to image/. Maya changed the line once, so the agent made a second card and kept the first. About 800 credits for both. |
| Website Agent | The Launch in 30 sales page link is dead. The Email Agent found it. Maya has not read that page since the last launch. | website agent. Check the Launch in 30 page. Tell me what is broken, then review the copy against the offer spec. | A fix list in order, the broken link at the top, then every line that says or implies the course closes. Maya says fix the top five. The agent asks the QA Agent to check the page before it goes live. |
| Sales Agent | About 20 DMs a week ask which one is right for me. Maya answers 14 of them, usually two days late. | sales agent. Here are five DMs from this week, word for word. Draft a reply for each in my voice, say which offer fits, do not send. | Five replies, each with the offer that fits and why. One sends a dog trainer with no email list to the workbook instead of the course, with a note to check that call. Maya approves all five and sends them herself from Instagram. About 540 credits. |
| Proof Agent | 214 students, 9 testimonials. Nobody asks. | proof agent. Build the proof bank from the nine testimonials in my Drive, sort by offer, and draft the permission ask for each. | A proof bank in Project files, nine entries, each with where it came from and whether permission exists. Nine permission drafts. Maya approves the asks. Every other agent can now pull from the bank, and the Delivery Agent’s day 60 ask feeds it. |
| Delivery Agent | Jordan Lee bought Launch in 30 on the weekend. The welcome goes out whenever Maya remembers, and nobody ever asks for a testimonial. | delivery agent. New student, Jordan Lee, bookkeeper in Ohio, paid in full Saturday, 10 module outline, wants to launch in January. Set up onboarding and the testimonial asks. | A delivery record, a welcome email with the next live Q&A date, and day 30 and day 60 testimonial asks, all drafted for the VA to send. It put the Q&A on the wrong week once. Maya said which Thursday. It fixed the line and saved all three to delivery/. About 450 credits. |
| Operations Agent | Maya opens the inbox every Monday and loses an hour. About 60 emails a day across support and partners. | Set up as an automation in module 15. Direct: operations agent. Here are the six things in my inbox from the weekend. Build Monday’s briefing. | Three need Maya today: a failed card at Kajabi, a workbook refund request, Tuesday’s thumbnail. The rest handled or parked, with a line for the VA. It refused to draft a reply on the money thread, which is right. Maya reads it in five minutes. About 330 credits. |
| Finance Agent | Maya watches the Stripe dashboard, not the books, and finds out about a bad month six weeks late. Ads cost $4,800 a month and nobody knows the cost per course sale. | finance agent. Revenue by offer this month versus last, refunds, ad spend, and cost per Launch in 30 sale. What could you not categorize? | A one page read from Stripe and PayPal, workbook versus course versus Room, and the transactions it could not place. Maya answers those. She puts it on a Friday automation. |
| Strategy Agent | End of the month. Maya is not sure whether to fix the sequence or add a third launch. | strategy agent. Score my six pillars from everything in this Project and tell me the constraint. | Six scores with the evidence for each, pulled from what the other agents saved this month, and one constraint: the evergreen path, not another launch. Maya leaves the third launch alone and puts the Sales Agent on a daily automation. |
Each agent, what to say to it
Foundation
Holds your profile, voice, offers, stories and brand. Every other agent reads it first.
Open itOpen its session only to update it. Say: foundation, update.
- Update my pricing to X and tell every agent.
- I changed my positioning, here is the new sentence, update the profile.
- Teach me what you know about my voice in five lines.
Offer
Designs and sharpens offers. Writes the spec the Funnel and Email agents build from.
Open itNew session, say: offer agent. Then the job.
- Review my main offer against my map and tell me what is weak.
- Design a low ticket front end for the people who are not ready for my core offer.
- Write the offer spec for X so the Funnel Agent can build it.
Funnel
Turns an offer spec into a funnel plan and builder handoffs. Asks for a credit ceiling before building anything.
Open itNew session, say: funnel agent. Give it the offer name.
- Plan the opt in funnel for X, pages, emails and the ad angle, nothing built yet.
- Build the application funnel for X with a 2,000 credit ceiling.
- Review my current funnel at this URL and tell me where people leave.
Writes and loads campaigns and sequences in your voice. Loads only with your approval.
Open itNew session, say: email agent.
- Write this week’s email on this idea, in my voice, and show me before loading.
- Write a five email welcome sequence for new subscribers from X.
- Read last month’s sends and tell me what worked.
Content
Plans and drafts content from your calls, videos and writing. Hands visuals to Image.
Open itNew session, say: content agent.
- Pull this week’s content from my last three calls and give me five post drafts.
- Turn this video transcript into a newsletter and three short posts.
- Plan next month’s content around the offer X.
Image
Renders brand visuals: quote cards, carousels, ad statics, in your brand card.
Open itUsually called by Content or Ads. Direct: say image agent.
- Make quote cards for these three lines in my brand.
- Render the carousel Content just drafted.
- Give me three ad statics for X at Meta sizes.
Website
Builds and revises pages from brand and offer.
Open itNew session, say: website agent.
- Build a one page sales page for X from the offer spec.
- Review my home page and list the ten fixes in order.
- Rewrite the about page in my voice.
Video
Scripts, cut lists and production handoffs.
Open itNew session, say: video agent.
- Script a three minute video on X for cold traffic.
- Turn this call into five short clips with timestamps.
- Plan a weekly video series around my core offer.
Ads
Creates, scores and optimizes Meta ads from your account.
Open itNew session, say: ads agent.
- Review the last 30 days and tell me what to kill, scale and refresh.
- Write six ad variations for X and score them before I see them.
- Set up a test campaign for X, paused, and show me the plan.
Writing
Long form, guides and books in your voice.
Open itNew session, say: writing agent.
- Outline a guide on X from my frameworks.
- Draft chapter one from this outline and my stories.
- Edit this piece for my voice without changing the argument.
Visibility
Podcasts, PR and getting seen.
Open itNew session, say: visibility agent.
- Find twenty podcasts my customer listens to and draft the pitch.
- Write my speaker one sheet from Foundation.
- Plan a visibility push for the launch of X.
Sales
Pipeline, follow up, DM replies, inbound triage.
Open itNew session, say: sales agent.
- Read my inbox and calendar and list who needs a follow up today, with drafts.
- Draft replies to these three DMs in my voice, do not send.
- Score these leads against my ideal customer.
Proof
Collects testimonials and tracks permission.
Open itNew session, say: proof agent.
- Find every testimonial in my Drive and calls and sort by offer.
- Draft the permission ask for these five clients.
- Build a proof bank for the sales page for X.
Delivery
Client onboarding, tracking and the work log.
Open itNew session, say: delivery agent.
- Set up onboarding for a new client in X.
- Read this week’s client messages and draft the replies for my yes or no.
- Show me every open promise to clients.
QA
Checks pages, funnels and deliverables before they ship.
Open itNew session, say: qa agent. Give it a URL or a file.
- Check this page on desktop and phone and list what is broken.
- Review this PDF before it goes to the client.
- Test this funnel end to end and report.
Operations
Daily operations, briefings, the inbox.
Open itBest as an automation, module 15. Direct: say operations agent.
- Give me today’s briefing: calendar, inbox, what is due.
- Triage my inbox and draft replies for the ones that need me.
- List everything I said I would do this week.
Intimacy
Everything after the sale: welcome, billing rescue, refunds and cancels, churn watch, reactivation. Drafts first, you approve every send.
Open itNew session, say: intimacy agent.
- A customer just bought X, set up their welcome and first week.
- Who has gone quiet in the last 30 days, and draft the check in for each.
- A card failed on X, draft the billing rescue email.
Outreach
Reaching people one at a time before they are a conversation. Warm, network and cold lanes, each with its own rules.
Open itNew session, say: outreach agent.
- Here is my list of twenty people, score them and tell me who to reach first.
- Write the opener and two follow ups for X in my voice.
- What do we know about X before I message them.
Strategy
Scores your six pillars and tells you what to fix first.
Open itNew session, say: strategy agent. Monthly.
- Score my six pillars from everything in this Project and tell me the constraint.
- Review my map against what the agents produced this month.
- What should I add next, and why.
Mindset
Mentor conversations that know your patterns. Tells you the cost before it starts.
Open itNew session, say: mindset agent.
- I am stuck on a decision, walk me through it.
- What pattern do you see in how I avoid X.
- Weekly check in.
Finance
Weekly transaction review, books sync, flags what it cannot resolve.
Open itBest as a weekly automation. Direct: say finance agent.
- Review this week’s transactions and tell me what you could not categorize.
- Revenue by offer this month versus last.
- Build a simple cash forecast for the next 90 days.
Before you ask
Do I have to use the exact names?
No. Describing the job works too. Names are just faster, and they stop two agents loading at once.
What if two agents should work together?
Ask the first one to hand off. Say: hand this to the email agent. Offer writes a spec Funnel reads, Content hands to Image and Email. Module 12 is all about this.
How do I know which agent to ask?
Follow the pillar. Marketing goes to Offer, Funnel, Email, Content, Image, Ads. Sales to Sales and Proof. Delivery to Delivery and QA. Operations to Operations. Sales after the sale to Intimacy, and before the conversation to Outreach. Finance to Finance. Not sure? Ask in the Project with no agent named and it will route.
Is Clear Launch Co. real?
No. I built it so I could run every agent on a business and break things without a client paying for it. Maya does not exist. Eight of the thirteen scenarios, and every screenshot, are what the agents actually produced in that demo Project. The other five are how those agents work, not a run.
How much does one of these scenarios cost?
Most of them land between 300 and 800 credits, which is $3 to $8. The first run of any agent costs more, because it sets itself up at the same time. The Email Agent’s first job in the demo was about 1,300. The Website Agent build is the expensive one, a few thousand. Module 17 has every number.
What if the output is generic?
Foundation is thin, or the agent did not read it. Ask: did you read the Foundation profile and the voice file. If no, say: read them and redo. If yes, add five of your real emails to Material and tell Foundation to update voice.
Can I give an agent a job that is not on my map?
Yes. The map is the order, not a fence. But if you find yourself giving agents a lot of jobs that are not on it, go back to module 8 and update the map. The map should match what you actually do.
How many agents should I be using in week one?
Two or three. The one on your top line every working day, and one or two more. Add one a week after that. Twelve agents in week one is how people burn credits and give up.
It is not using my voice or my Foundation answers. Do I have to tell it every time?
Name the skill in the ask, yes. Once you have more than a couple of skills it will not guess which one you meant, and the lighter models guess least. Write in my voice, use the voice skill. Ten extra words. The agents in the bundle do this for you, a plain session does not.
Do I need to know Meta Ads Manager to use the Ads agent?
No. Once Meta is connected you do not go into the Ads Manager screen again. You do need a Business Manager with you as admin on the ad account, which takes twenty minutes if you have only ever boosted posts. Campaigns upload paused. You look, you approve, then they run.
Can I ask an agent what it does before I use it?
Yes, and you should. Open it and type: what do you do and what do you need from me. It answers in a few lines. Cheaper than guessing from the name.