AI assisted software delivery
summary
Generative AI promises 20–50% faster coding but code writing is only about 26–27% of software work; experiments with GitHub Copilot (~500 licenses) and Tabnine show real gains for mid/senior engineers, less for juniors. Early lab work targets mainframe modernization and code-to-doc summaries, but hallucinations and IP leakage appear and the prediction is an initial surge of low-quality, hard-to-maintain code. Strategy needs tooling, governance, and preserving pair programming and developer expertise to maintain flow and avoid more legacy debt.
AI-written summary · about
chapters (10)
- 0:00The disruptive power of AI
- 1:45Strategy in an evolving space
- 4:02Beyond coding: The real work
- 7:31Radar and modern complexity
- 10:29Pairing and quality design
- 13:29AI as a development partner
- 17:28Improving the full lifecycle
- 21:59Expertise and future risks
- 26:14Lean principles and flow
- 30:46Governance in the AI era
related talks
51mAI-Assisted Coding: The Future; Between Challenges And Benefits
27mAI for software development: A reality check
44mAI Code Generation - Benefits, Risks and Mitigation Controls
27mSoftware engineering with GenAI
30mThe AI Revolution in Coding
19mWith great power comes great responsibility in code quality
01h 12mAI-Assisted Programming Today and Tomorrow
18mCoding Reinvented: How AI Agents Are Shaping the Future of Software Development