The Situation
Developers were repeatedly asking the same foundational questions across support channels, often for issues that already had clear answers. Support teams struggled to keep up, simple issues required escalation, and documentation was frequently stale or fragmented. At the same time, developers expected a broader source of truth—covering not just how to build, but also business context, operational processes, and launch cycles—without exposing internal or confidential information to external audiences.
The Strategy
I approached the problem with a core framing: the content itself was the product. Rather than adding more tooling, the focus was on increasing the quality, coverage, and accessibility of knowledge. I identified content gaps through developer interviews, support patterns, and unanswered queries, and used jobs-to-be-done to prioritize what information mattered most.
To deliver relevance without noise, I designed a just-in-time context model aligned to key signposts across the developer journey. This included clear rules for when content should appear, how deep it should go, and how it should adapt across internal and external developer audiences. A gating framework ensured sensitive information stayed internal while maintaining a consistent source of truth.
Finally, I treated information quality as a systems lever. By improving accuracy, freshness, and contextual clarity, the agent aimed to reduce downstream defects, sub-optimal implementations, and avoidable support escalation.
The Results
- Introduced telemetry concepts to measure usage, task completion, and DX outcomes
- Reduced repetitive support questions by surfacing contextual answers earlier
- Validated JTBD-specific agent personalities aligned to task intent
- Established groundwork for staged MCP integrations across internal and external agents
- Demonstrated opportunities to accelerate development velocity through better context
- Reduced manual support effort by externalizing high-confidence knowledge