← Kenrick Tan / Back to work

Case study

Internal AI Tooling Suite

Custom MCP server integrating LLM automation into Jira and Confluence workflows.

Built an MVP internal AI tooling suite that brings LLM-powered automation to regulated financial workflows.

Impact

45% faster Jira ticket resolution

Impact

62% reduction in documentation time

Impact

8 business lines adopted in 3 months

Impact

Zero data leakage in regulated environment

Problem

Knowledge workers across the enterprise spent hours creating, updating, and searching for documentation in Jira and Confluence, with inconsistent quality and discoverability.

Architecture

AI Tooling Architecture

Select a component to explore its role and connections.

123456

User

Engineers, Product, PMs

Connects to: LLM Gateway

Approach

  • Designed a custom MCP server exposing Jira and Confluence APIs as LLM tools.
  • Implemented scoped access controls ensuring data never leaves regulated boundaries.
  • Created specialized prompts for ticket summarization, documentation generation, and search augmentation.
  • Deployed within enterprise guardrails with audit logging and approval workflows.

Outcomes

  • Teams use AI to draft tickets, generate release notes, and create runbooks faster.
  • Knowledge is more discoverable with semantic search across Jira tickets.
  • Reduced cognitive load on engineers during release management cycles.
  • Security team approved deployment within regulated financial environment.

Metrics

Before → After comparison

Before

ticket Creation Time
18 minutes
doc Creation Time
45 minutes
search Success
58%
weekly Ai Tickets
0

After

ticket Creation Time
8 minutes
doc Creation Time
17 minutes
search Success
84%
weekly Ai Tickets
340+

Stack

TypeScriptMCP (Model Context Protocol)Jira APIConfluence APILangChain