An AI DM setter that books calls the way your best closer would
The traffic is already paid for. Whether the first reply lands, whether anyone follows up, whether anyone remembers by day three — too much depends on luck.
From first message to renewal — five places it leaks
In a knowledge business the sale is not one moment; it is a chain that runs for weeks. People drop out at every step, often not the low-intent ones.
0101 First DM
Gone quiet
Contacts you paid for with ads and live streams go quiet the moment the first reply misses.
0202 Booking
Empty webinar seats
The window from a cheap intro course to the full one is days. Miss the one-to-one, miss the window.
0303 Follow-up
After “let me think”
“Waiting on payday”, “asking my partner” — the biggest group, and the least likely to be chased after the third try.
0404 Show-up
Enrolled, no-show
Attendance drives completion, completion drives renewal. Everyone knows to send reminders; nobody sends them all.
0505 Renewal
Alumni left alone
Advanced programmes live on past students, but consultants’ hours go to new leads.
The four stages, in a course business
The four stages do not change. What changes is what each one is built around — your course lines, your judgement, your red lines, your takeover limits.
Codify the playbook
Your conversion path, not a script library
The stretch from intro course to full programme — the questions, the calls, the pace you push at, written into how your AI workers work.
Scale the execution
Roles drawn around your cohorts and course lines
First contact, booking, multi-day follow-up and course Q&A each get a role and limits, dispatched by a lead agent. Follow-ups pick up from last time, not from scratch.
Compound the learning
“A good conversation”, in your own assertions
Before launch the review workbench checks it: required lines must appear, forbidden promises must not — no pass, no deployment. Problem conversations later return as new cases.
Humans where they matter
High ticket, refunds, angry customers — your call
When a rule fires, the conversation goes to a person backstage; the customer still sees one name. Your consultants step in where they matter most.
What keeps it from saying the wrong thing
Put the four stages into runtime and you get five hard constraints — the shape of the product, not adjectives about it.
Entry conditions
Every stage checks its goal and move-on conditions first; unmet, it stays there. The push is a judgement, not a script.
No criteria, no launch
The review workbench runs first: deterministic assertions plus a judge model. A case with no way to be judged does not get through.
Knowledge by worker
Not one shared library for everyone. The intro course and high-ticket programme each get their own; change the line, change what it can quote.
Contact-level takeover
The switch sits on the contact, not the account: this one goes to a person, the rest keep running, and the customer still sees one name.
Turn-by-turn record
Reasoning, tool calls and timings are recorded turn by turn. The broken turn becomes a case the next version must pass.
The part you do not hand to AI
Here the expensive mistake is not a lost deal but a sentence that should never have been said. So boundaries come in two kinds: some go to a person, some become checks.
Goes to a person
- Refunds, complaints：Money and emotion go to a person; the AI only hands over a clean context.
- Outcome promises：Whether they will earn it back or pass — nobody should promise that, an AI least of all.
- Big and one-off：High-ticket coaching, one-to-one and corporate deals go to a person from the first line.
Written as a rule
- Forbidden lines：Phrases like “guaranteed pass” become assertions that must never fire, and sensitive-topic limits sit here too — no pass, no launch.
- Must-says：Refund terms, course length, credentials — what must be said gets checked.
- Handoff triggers：Amount, keyword, tone — any one of them hands the conversation back.
What may be claimed about outcomes and earnings differs by market — you write the rules, the product turns them into pre-launch checks.
Three questions course businesses ask
We already run a CRM and messaging tools. Why add this?
They do different jobs. A CRM holds people and process — who sits in which pool, what gets sent when. TOPPP runs the conversation itself: reading intent, taking the objection, deciding the next move, along your proven path. With both in place, the people in the pool stop just sitting there. And it can connect with your CRM.
Our playbook changes every cohort. Can it keep up?
Configuration is versioned: save a new version, pass the review, then apply it. Until you apply, the workers already running are untouched — and you can roll back. Once applied, the shared part of a fix lands across the whole team at once, and conversations already in flight switch over on the spot. Changing between cohorts need not mean breaking between cohorts.
We run several course lines. Can each keep its own playbook?
Yes. Each line is its own project — own workers, knowledge base, review criteria and takeover limits. An intro course and a high-ticket programme should not ask the same questions.
Bring this cohort’s playbook — watch it become a team
A demo takes about 30 minutes and covers three things:
- 01Play a prospective student and watch the AI worker handle opt-in, booking and follow-up
- 02See the agents’ routing, decisions and human handoff boundaries in real time
- 03Run both ledgers on your own opt-in rate, reply time, booking rate and follow-up reply rate
Or reach us directly business@toppp.ai