Case study - Sanskrit coaching-OS
A full operating system (OS) to run the “Teach Sanskrit to kids” project using agentic AI plus human-in-the-loop execution.
0. System overview
Architecture
[Traffic] → [Acquisition Agents] → [Funnel System]
↓
[Conversion Agents]
↓
[Delivery System]
↓
[Retention Agents]
↓
[Growth Engine]
Core idea
Humans focus on teaching and quality. Agents handle marketing, analysis, operations, and scaling.
1. Agent stack (core agents)
Acquisition agent
Role: Drive traffic and leads.
Tasks
- Generate ad creatives (reels, copy)
- Suggest targeting
- Run A/B tests
Tools
- Meta Ads
- Google Ads
- Canva / CapCut
Input
- ICP (parents, kids of learning age)
- Messaging pillars
Output
- Ads
- Campaign performance insights
Example prompt
ROLE: Acquisition Agent
TASK:
Generate 5 Instagram ad creatives for Sanskrit learning for kids.
Focus on cognitive benefits + engagement.
OUTPUT:
- Hook
- Script
- Visual idea
Funnel agent
Role: Optimize landing page and conversion.
Tasks
- Analyze drop-offs
- Suggest improvements
- Run A/B tests
Tools
- Framer
- Google Analytics
- Hotjar
Output
- Conversion insights
- UI/UX improvements
Sales agent (WhatsApp)
Role: Convert leads into paying customers.
Tasks
- Answer parent queries
- Handle objections
- Push trial to paid
Tools
- WhatsApp API
- CRM (HubSpot / Airtable)
Example script
Parent: What will my child learn?
Agent:
Great question! In just 4 weeks, your child will:
- Speak simple Sanskrit sentences
- Chant 5–7 shlokas confidently
- Improve memory and pronunciation
Would you like to try a free session this weekend?
Decision agent
Role: Evaluate business health.
Tasks
- Monitor CAC, conversion, and retention
- Recommend scale, iterate, or stop
Output
- Weekly decision report
Curriculum agent
Role: Design teaching content.
Tasks
- Create lesson plans
- Design games, exercises, and chants
Output
- Weekly curriculum
- Worksheets
Engagement agent
Role: Keep kids engaged.
Tasks
- Create quizzes and games
- Suggest interactive activities
Retention agent
Role: Improve renewals.
Tasks
- Send progress reports and reminders
- Trigger renewal nudges
Operations agent
Role: Run backend operations.
Tasks
- Create SOPs
- Manage schedules
- Assign instructors
Hiring agent
Role: Scale team.
Tasks
- Define roles
- Screen candidates
- Generate interview questions
2. Tools stack
Core stack
| Function | Tool |
|---|---|
| Website | Framer |
| Payments | Razorpay / Stripe |
| CRM | Airtable / HubSpot |
| Communication | WhatsApp API |
| Analytics | Google Analytics |
| Ads | Meta / Google |
| Content | Canva / CapCut |
AI stack
- LLM (Claude / GPT)
- Vector DB (optional later)
- Automation: Zapier / Make
3. Workflow (end-to-end)
Flow 1: Acquisition to conversion
Ad → Landing Page → Lead Capture → WhatsApp → Trial → Payment
- Acquisition agent launches ads
- Funnel agent monitors performance
- Lead captured (form)
- Sales agent initiates WhatsApp
- Trial booked
- Parent converted
Flow 2: Delivery
Enrollment → Batch Assignment → Classes → Feedback
- Operations agent assigns batch
- Curriculum agent provides lesson plan
- Instructor delivers class
- Engagement agent enhances interaction
Flow 3: Retention
Class → Progress → Parent Update → Renewal
- Engagement agent tracks participation
- Retention agent sends updates
- Renewal nudges sent
Flow 4: Growth loop
Great Experience → Referrals → New Leads → Scale
4. Metrics dashboard
Acquisition
- CTR
- Cost per lead
Conversion
- Trial to paid percentage
- CAC
Retention
- Renewal rate
- Drop-off rate
Business
- Revenue
- Instructor utilization
5. Human and AI role split
Humans
- Teaching
- Quality control
- Strategy
Agents
- Marketing
- Optimization
- Operations
- Analytics
6. Scaling playbook
Stage 1 (0–50 students)
- Manual execution plus basic agents
Stage 2 (50–200 students)
- Hire instructors
- Automate sales and retention
Stage 3 (200+ students)
- Build product (app)
- Full agent orchestration
7. Advanced
Multi-agent system: LangGraph / CrewAI, where each agent is a module.
Example
User Lead →
Acquisition Agent →
Sales Agent →
Decision Agent →
Ops Agent →
Retention Agent
Final insight
This becomes a self-improving system that learns from every user interaction.