Streamlining business analysis with a Claude-powered knowledge base

• TriTech's SDLC business rules are complex and tightly interconnected, and they vary across all 10 U.S. states the product supports, so a single requirement change could ripple into unrelated functionality without warning.
• Gaps and edge cases in requirements sometimes weren't caught until development or QA, when fixing them meant costly rework instead of a quick spec update.
• As requirements evolved mid-implementation, keeping documentation, stakeholder alignment, and cross-references consistent became a slow, manual task.
• Business analysts had no fast way to check a new requirement against a large, growing body of existing specifications, making it hard to balance business intent against technical constraints with confidence.

Proxet's BA team, running the requirements and analysis phase of the product lifecycle for TriTech, redesigned that workflow around a Claude-powered knowledge base built directly into day-to-day documentation work.

• Structured product requirements, business rules, state-specific logic, and product knowledge into a file-based knowledge base that Claude could search, reason over, and update alongside the team's existing tools.
• Used Claude Code to run impact analysis on proposed changes, surfacing potentially affected functionality, business rules, and states before a change was approved.
• Partnered with QA to generate structured test cases with Claude, including edge cases and "unhappy path" scenarios, directly from requirements and user stories.
• Cross-checked new requirements against the specification set to flag conflicts and inconsistencies before they were finalized, rather than after implementation began.

• Reduced review time: from hours to minutes. Manually tracing whether a new business rule affects the relevant reports, states, and state-specific exceptions could take an estimated 1–2 hours of review. With the knowledge base surfacing relevant files and cross-references, the same analysis took approximately 5–10 minutes.
• QA support: For a single reporting feature, Claude generated 50 structured test cases across 13 test categories, including edge cases, negative scenarios, and cross-feature impact checks, directly from the requirements and acceptance criteria.
• Moved requirement-conflict and edge-case detection earlier in the lifecycle, ahead of development and QA instead of after code was already written.
• Validated the approach on real project work — business rule lookup, impact analysis, requirement conflict detection, and QA test case generation — and completed the handoff to TriTech's team leadership.