The Future of Software: AI Copilots & Developer Rethinking
Inbal S
Chief Product Officer, GitHub
FORMERLY: AWS, MICROSOFT
The Shift
Developers: From Code Writing to Systems Thinking
"Developers need to form a different thinking. You start figuring out how to use AI tools to be successful. It's no longer just code writing—it's evolving your thinking to the big picture, to connected systems."
Junior devs can spend time understanding systems from day one instead of learning syntax
Architecture and design thinking move earlier in developer careers
The human in the loop remains central—Copilot is a copilot, not a pilot
92% of developers already using AI tools—it's table stakes
Accenture: 88% of suggested code retained in production
Developers spend <25% of time actually writing code; rest on meetings, reviews, builds, legacy code
What's Most Underhyped
AI-driven testing: More code → more critical to test. Unit, integration, load, security, penetration testing all need AI-powered generation.
What's Overhyped
Replacement: AI won't replace humans. You need humans in the loop because AI cannot replace innovation, that creative spark that is the center of humanity.
The Mistakes Teams Make
How Companies Get AI Adoption Wrong
Expecting magic: "Here's a tool, go use it." Requires change management, not just tool deployment
AI for AI's sake: Asking "What should we do with AI?" instead of "What problem are we solving?"
Starting backwards: Plastering AI on things instead of identifying workflow friction first, then applying AI
Wrong metrics: Time is not quantifiable—you can write bad code fast. Focus on time-to-value instead
The Right Framework
Start with customer problem → identify where developers lack time (meetings, reviews, legacy code digging) → apply AI to automate manual/config-heavy tasks → measure time-to-value and developer happiness
Key Insight
Developers don't lose jobs. They shift to higher-value work: collaboration, innovation, architecture. That time gain = happier devs = better retention.
Metrics That Matter
Measuring Copilot Success
Not just time: Translate time → efficiency → productivity
Code quality: Can you improve security with Advanced Security tools?
Developer happiness: The ultimate metric—are devs more focused, less frustrated?
Time-to-value: From task assignment to realized value (revenue, adoption, market speed)
The TrapProductivity is not the right metric for every use case. For security tools, measure prevention (secrets leaked, bugs caught). For collaboration, measure cycle time. Combine metrics = developer happiness.
Contrarian Truths
What Inbal Gets Right About AI in Dev
✗AI replaces developersINSTEAD →✓ AI frees devs from toil. They do more creative, high-value work. Retention improves.
✗25% efficiency = 25% fewer peopleINSTEAD →✓ Dev time unlocked goes to collaboration, innovation, architecture. You need the humans more, not less.
✗More code = more productivityINSTEAD →✓ You can write bad code fast. Productivity = time-to-value + quality + developer happiness combined.
✗Copilot is a replacement toolINSTEAD →✓ Copilot is a collaboration and leverage tool. Devs pick and choose what to delegate, what to keep crafting.