AI Tools for Engineering and Product Productivity Improvement

We’ve heard a number of CTOs ask some version of this question:

I have a directive from my CEO to use AI to achieve 20-30% productivity improvement across product and engineering in 2025.

I need to present a plan in 30 days and implement the plan in 90 days. What should I do?

I think a lot of us have a similar question. We have some adoption of AI tools, but it’s uneven. How do I think about getting appropriate adoption in my organization?

This is going to be a structured conversation where LA CTO Forum members will share their experiences and advice around AI planning and implementation including:
Areas to Consider
1. Code Generation/AI-assisted programming
-Discuss tools like Cursor, GitHub Copilot, Tabnine, Augment code, AI copy/paste, manicode/codebuff, etc
-Automate repetitive coding tasks
-Generate code that follows existing patterns in a codebase
-Refactoring code / Reducing / Removing tech debt
-Translating code to other languages
-Documentation generation
2. Software Quality Assurance with AI
-Automated test case generation
-Bug and vulnerability detection
-Automated code reviews
-Performance optimization suggestions
3. Product Planning/Product Exploration/Product Definition?/Roadmap
-AI driven analysis of customer feedback/tickets/usage data to discover holes in desired functionality/most hated bugs
-Using AI for product innovation
-AI-enhanced UI/UX development
-Streamlining Product Definition Processes
-Enhancing Roadmapping and Strategy
-Clustering user data to create detailed user personas
-Risk analysis

Getting This to Happen
1. Technical concerns
-Loss of IP
-Data handling risks
-Non-programmers generating code
-Generation of vulnerable/low quality code (can then tie into topic 2)
-Adversarial attacks against models themselves
2. Non technical concerns
-Finding internal champions to advocate for AI
-How to measure improvements?
-Skill degradation/dilution due to overuse
-Morale considerations / dealing with pushback from devs?
-Licensing/Attribution challenges