Engine

Summary

I designed and implemented an end-to-end automation that took people from cold Instagram interaction to paid community member, without the founder touching DMs, manual interviews, or payment coordination. AI handled most of the qualification; edge cases were escalated to humans.

Context

Engine is a community-driven business. The founder was stuck in:

• Answering cold DMs on Instagram
• Manually vetting whether someone was a fit
• Sending invites and links by hand
• Chasing payments
• Notifying the team to onboard new members

It worked at low scale, but it was eating his entire day and blocking growth.

Problem & constraints

Business objective:

Free the founder from day-to-day intake and onboarding while keeping the quality bar for new members.

Constraints:

• Must start from Instagram DMs (where leads already came from).
• Must classify candidates against an “ideal member” profile, not just check a form.
• Must keep a human in the loop for edge cases.
• Must handle invite, payment, and onboarding automatically once someone is accepted.

System design

I mapped the intake as a state machine:

1. New lead – cold IG DM starts a flow.
2. Survey sent – bot collects answers to key questions.
3. AI classification – responses are evaluated against the ideal profile.
4. Decision routing:
   • Approved → continue automatically.
   • Unclear → send to human for review.
5. Event invite – approved candidates receive a QR code for the event.
6. Attendance – QR scanned at the door to confirm presence.
7. Payment – after the event, approved attendees receive a payment link.
8. Onboarding – once paid, internal team is notified and the member is introduced properly.

Each stage and transition was tracked in a lightweight CRM.

Implementation & my role

I owned the design and build of the full workflow:

• Conversation & entry:

  • Instagram automation and message handling via ManyChat.
  • First survey to collect candidate information.

• AI-assisted decision:

  • OpenAI API classifying responses against an “ideal member” profile.
  • Automatic “approve / needs human review” routing.

• Pipeline & tracking:

  • CRM/pipeline built in Airtable to track each lead’s state.

• Invites & payments:

  • QR code generation and delivery for event entry.
  • Timed follow-up after attendance with Stripe payment link.

• Onboarding:

  • Post-payment survey to capture details.
  • Slack notification to the team when a new member joined so they could prepare a personalized introduction.

• Analytics:

  • Basic analytics to count visits and follow the funnel, at the founder’s request.

The entire workflow was designed around real daily operations of the community and my own experience as a member.

Results

• The founder no longer had to live in Instagram DMs or manually chase every step.
• Candidate fit was evaluated consistently by AI, with human review only for edge cases.
• Invites, attendance, payments and onboarding happened on rails once someone was approved.
• The team had clear visibility into who was coming, who paid, and who needed a welcome.

This moved intake from a personality-driven process to a repeatable system.

Key learning

Two big takeaways:

1. Automation is workflow design, not tools. The success of this project came from mapping the real-world states (DM → survey → AI decision → invite → payment → onboarding), not from any specific platform.

2. AI works best as a decision component, not “the whole system.” Using the model for classification and routing, with humans on unclear cases, gave us leverage without losing control.