Scaling a Career Coaching Platform from 15 to 100 Participants
Replacing manual coaching processes with 6 conversational AI agents for Zinc VC's Inflection Programme.
The Challenge
Zinc VC runs the Inflection Programme, a fellowship for experienced mid-career professionals looking to redirect their careers toward mission-driven impact. The programme had been running successfully with around 15 participants, relying on manual coaching sessions and administrative processes.
As demand grew, Zinc wanted to scale to 100 participants — but the existing manual approach simply wouldn't work at that volume. They needed to maintain the quality and personalisation of the coaching experience while dramatically reducing the per-participant cost and effort.
The Approach
Rather than building a monolithic platform, I designed a system of specialised conversational AI agents — each handling a distinct part of the coaching workflow. This agent-based architecture meant each component could be developed, tested, and refined independently.
All 6 agents were built in N8N, alongside an orchestration workflow that routed incoming messages to the right agent. N8N's visual workflow builder meant Zinc's team could understand and modify agent behaviours without needing development support.
What Was Built
6 AI agents specific to each stage of the programme
Annotation dashboard to review conversation traces and evaluate efficacy of the system
Embedded into Circle's authentication system
Conversation interface embedded within the Circle platform
Recommendation engine for suggesting relevant resources based on participant progress
Technical Architecture
- N8N
All 6 agents and the orchestration workflow built here. Visual debugging and easy modification by non-technical team members.
- Airtable
Backend data store for all agents. Already familiar to the team — no migration needed.
- Circle API
Agents posted on participants' behalf to Zinc's existing community platform.
- Netlify
Hosted the annotation dashboard — a web app for exploring conversation traces and analysing agent responses.