This page summaries approaches for improving procceses in IT company and in its products via AI/LLM.

Possible chalenges

• Low/resource consuming unit test coverage
• E2E gaps in automated testing or high time/resource cost for such coverage
  Time to market!
• Lost bug fixes between branches
• XXXk jira tickets → difficult to reuse knowledge
• Slow resolution time for recurring bugs

AI Opportunities

• Code → AI-generated tests, reviews, branch fix detection
• Testing → automated test case generation, prioritization
• Tickets → smart search, duplication detection, fix recommendations
• Operations → anomaly detection, predictive maintenance
• Support → intelligent routing, knowledge assistant
• Product — changes in core engine of product to use LLM for smart behaviour adaptation

AI in Action (Examples)

• AI generates JUnit tests for uncovered classes (saves 30% dev time).
• Bug triage assistant finds 90% similar tickets → avoids duplicate fixes.
• ML-powered CI pipeline runs only most relevant tests → CI time cut by 40%.
• Branch fix recommender highlights missing bug fixes in other branches.

Benefits

• Faster delivery (shorter release cycles)
• Higher code quality (fewer production bugs)
• Lower cost of testing & maintenance
• Better knowledge reuse (faster bug fixing)
• Happier customers (fewer outages, faster fixes)

Next Steps

1. Pilot project:
   • AI for Jira ticket deduplication + knowledge assistant
   • AI-assisted unit test generation
2. Measure impact (coverage ↑, resolution time ↓)
3. Scale successful pilots to whole company

LLM cost

LLM-selfhosted (office) vs LLM-selfhosted (hosting) vs LLM via API (openai and others) vs hybrid (LLM-selfhosted for simple inputs and more advanced tasks via LLM via API)

Final goal for AI transformation expert: You’ll want to keep it practical and show ROI: “if we reduce bug resolution by 20%, we save X dev hours = Y $$ per year.”

Task1

Generating unit tests for Java with mocking
Article