Overview of Leaders of Code — “When the cost of code approaches zero, what does engineering leadership look like?”
In this episode of the Stack Overflow podcast’s Leaders of Code segment, Ira May and Ben Matthews speak with Eric Anderson, Director of Engineering at Intuit, about how AI is reshaping software development, engineering leadership, and team collaboration. The conversation centers on a major shift: when generating code becomes nearly instantaneous, the real bottlenecks move from writing code to ideation, design, experimentation, and human judgment.
Key Themes and Takeaways
AI makes code cheaper, but judgment more valuable
- Eric argues that the “incremental cost of a line of code” is now extremely low.
- Because code generation is faster, engineering leaders must rethink:
- what it means to build a feature
- how system design works
- how teams measure effectiveness
- The human role becomes even more important in deciding what to build and why it matters.
Customer value remains the core metric
- Traditional metrics like PR counts, reviews, and lines of code still exist, but they’re not the most important.
- The real question is: Did the work create customer value?
- Intuit uses experimentation heavily to compare product experiences and improve outcomes faster.
The bottleneck has shifted upstream
- The main slowdown is no longer code production itself.
- The new bottleneck is the path from idea → design → requirements → implementation → production.
- Eric says teams need to retool their processes, especially around:
- design iteration
- product requirements
- handoffs between product, design, and engineering
- defining what “done” means in an AI-accelerated workflow
Roles are blending across disciplines
- Eric describes a more fluid boundary between:
- product management
- design
- engineering
- data science
- PMs can now generate PRs and prototypes more easily.
- Engineers are expected to understand product thinking more deeply.
- This creates stronger cross-functional empathy and faster decision-making.
Engineering Leadership in the AI Era
Leaders need to manage systems, not just code
- Engineering leadership is becoming more like orchestration:
- managing people
- managing agents
- managing feedback loops
- ensuring quality, maintainability, and operational excellence
- Even if AI can generate code quickly, teams still own:
- CI/CD pipelines
- deployment
- observability
- supportability
- long-term maintainability
Senior engineers gain leverage faster
- Senior staff engineers at Intuit are reportedly much more productive with AI tools because they already understand architecture and system structure.
- Their experience helps them use AI effectively without losing control of quality.
Junior engineers need more support
- Eric raises concern that junior engineers may struggle if they rely too heavily on generated code without understanding it.
- He emphasizes that software craft used to be learned by working closely with more experienced engineers.
- The industry now needs new ways to teach:
- ownership
- debugging
- modular design
- algorithmic thinking
- responsible use of AI tools
What Still Matters for Engineers
Core skills are still foundational
Eric says strong engineers still need to understand:
- Algorithms and problem solving
- especially for data pipelines, ETL, and large-scale systems
- Modularity and structure
- breaking large problems into smaller components
- Ownership
- being accountable when things go wrong
- Customer obsession
- understanding who the software is for and what problem it solves
- Speed as a habit
- being able to unblock yourself and move forward quickly
AI doesn’t replace understanding
- AI tools can accelerate work, but they do not remove the need to understand how software works.
- Eric warns that a tool can generate output, but the engineer still owns the result.
- The deeper risk is not just bad code — it’s people losing the ability to reason about the code at all.
How Eric Uses AI in His Own Work
Practical uses beyond code generation
Eric says he uses AI for many task-oriented parts of his day, including:
- summarizing and drafting email responses
- triaging Slack messages
- synthesizing specs and technical docs
- comparing PRDs with code to check whether the intended outcome was built
- analyzing logs and diagnosing anomalies
- supporting research for product and competitive analysis
- helping draft promotion documents by aggregating evidence from projects, code, and documentation
Human-in-the-loop is still essential
- He originally tried to let AI send emails for him, but quickly stopped.
- His preferred model is:
- AI summarizes, organizes, and suggests
- humans review, decide, and finalize
- This approach works especially well for unstructured information and high-volume context.
AI Agents and the Future of Work
Agents are becoming part of the engineering process
- Eric describes a future where engineers manage specialized agents for tasks like:
- bug-finding
- code fixing
- quality improvement
- This creates a new management layer:
- monitoring what agents are doing
- checking whether they’re producing useful outcomes
- preventing them from going off-track
New risks require new controls
- Agents can optimize for the wrong thing if left unchecked.
- Eric suggests organizations may need “adversarial” agents to monitor other agents.
- This is part of a broader shift toward managing autonomous workflows rather than just writing code.
Career Advice for Engineers
Differentiate on how you work, not just what you know
Eric breaks engineering evaluation into two buckets:
-
Functional skills
- algorithms
- problem solving
- modular design
- debugging
- CI/CD understanding
-
How you do work
- ownership
- customer focus
- speed
- ability to unblock yourself
- depth of thinking
What stands out most now
He suggests engineers should focus on being:
- strong problem solvers
- adaptable
- collaborative
- customer-oriented
- comfortable with ambiguity
Final Takeaway
The episode’s core message is that AI is not eliminating the need for engineers — it’s changing the nature of their value. As code becomes cheaper to produce, leadership is less about maximizing output and more about guiding judgment, collaboration, experimentation, and accountability. The future engineer is part builder, part systems thinker, part product partner, and part AI manager.