The same questions — for 42 days straight.
Participants in a 42-day program kept asking about the day, phase and dosing — answers had to be context-aware, multilingual and compliant.
We built a RAG bot grounded in the program methodology, on Mastra and Claude. It knows which day and phase a participant is on and answers for that exact stage, only within the approved materials; sensitive questions are not improvised but escalated to a live coach together with the context. It answers in the participant's language on the web, in Telegram and in WhatsApp.
Before, routine questions went to coaches, an answer depended on a coach being free, and there was a real risk of answers drifting outside the methodology. Now the bot closes the routine within the methodology around the clock in three channels, the guardrails keep it inside the approved base, and coaches spend their time on coaching rather than on "which day am I on".
What was in the way
- ✗Participants asked the same things daily: which day, which phase, what to take
- ✗Answers had to stay within the methodology — no improvisation
- ✗Coaches spent time on routine questions instead of coaching
What the system does
Methodology-grounded answers
RAG over the program base — the bot answers only within approved materials.
Day and phase context
Knows which day of the program the participant is on and answers for that exact stage.
Guardrails and escalation
Sensitive questions are not improvised — the conversation goes to a live coach.
Multilingual
Answers in the participant's language on web, Telegram and WhatsApp.
How we roll it out.
Requests and access analysis
We study historical requests, verify CRM and API access, and score the automation potential.
A slice of real traffic
The agent runs on part of the live requests with weekly syncs — real dialogues, not a demo.
Metrics and scale-up
We evaluate volume, response speed and escalation share — then expand the agent to full traffic.