How I identified a documentation retrieval gap across engineering teams, designed and led the adoption of Enjo AI, and built a phased rollout roadmap from scratch, cutting search time and reducing cross-team interruptions.
My Role Technical Writer
Tools Enjo AI, Confluence, Slack, Google Drive
Organisation BookMyShow (Mumbai)
Timeline May 2025 – ongoing
<aside> 💡
90%+ accuracy threshold set for the agent before full rollout, tested across 100+ queries
</aside>
<aside> 💡
3 sources Confluence, Google Drive & Slack, all connected and indexed into one AI interface
</aside>
<aside> 💡
5 phases structured rollout roadmap I authored, from setup to continuous improvement loop
</aside>
<aside> ℹ️
Note on metrics: Precise before/after time data was not formally tracked. Qualitative feedback from engineers confirmed a meaningful reduction in time spent searching for documentation and a decrease in repetitive Slack queries. Specific query volumes and adoption timelines are estimates based on observation.
</aside>
BookMyShow's engineering and PMO teams had no shortage of documentation: onboarding guides, RCA reports, process documents, and engineering specs were all stored across Confluence spaces and Google Drive. The problem was retrieval.
Engineers who needed a specific answer say the buffer percentage in a budgeting process, or the owner of an approval step, had to manually search Confluence keyword by keyword, skim multiple pages, or fall back to asking in Slack and waiting for a reply. New team members faced this friction most acutely, spending significantly more time on lookups than experienced colleagues.
"The knowledge existed. The problem was that it was invisible at the moment of need." — Observed pattern from engineering team interactions
This created two compounding costs: time lost to individual searches, and senior engineers repeatedly interrupted to answer questions already documented somewhere. Identified this gap in my day-to-day work as the embedded technical writer.
Before proposing any solution, I spent time understanding how people actually searched for information and where they got stuck.
<aside>
What I observed
Engineers resorted to Slack when Confluence searches failed. Questions like "what's the buffer percentage?" or "who approves this?" were asked repeatedly, despite written answers existing in documentation.
</aside>
<aside>
Root cause identified
Confluence's keyword search required knowing the exact page name or terminology. It didn't understand questions in natural language, making it invisible to engineers mid-task.
</aside>
<aside>
Scope of documentation
Coverage included onboarding docs, RCA reports, PMO process documents, engineering specs, and historical Slack Q&A threads, all siloed across different platforms.
</aside>
<aside>
Key insight
The fix wasn't writing more documentation. It was making existing documentation answerable and queryable in natural language, with source citations, in the tool engineers already used: Slack.
</aside>
I proposed, evaluated, and led the adoption of Enjo AI, an AI agent platform by Troopr, which connected directly to our Confluence spaces, Google Drive folders, and Slack channels to create a single, queryable knowledge interface. I authored a full phased rollout roadmap for the implementation.