Building the most AI-pilled engineering team in the world | Fiona Fung (Manager of the Claude Code and Cowork Teams)

Jun 21, 2026 Episode Page ↗
Overview

Fiona Fung, Anthropic engineering leader, discusses how AI is transforming software engineering, with her team shipping 8x more code. She shares insights on hiring, team culture, managing high velocity, and the future of engineering roles and planning.

At a Glance
18 Insights
1h 38m Duration
24 Topics
6 Concepts

Deep Dive Analysis

Introduction to Fiona Fung and AI's impact on engineering

Evolution of the engineering role over 25 years

An AI-native software team in 2026

Using Claude for management and team output review

Shifting focus to verification and automated code review

Hiring profiles for AI-native teams: creative builders and deep systems experts

The mindset shift to ambitious thinking with AI

Thriving in AI-native teams: growth mindset and facing fear

Helping small businesses adopt AI tools

Anthropic's approach to spotting latent demand

The next frontier: asynchronous work with AI routines

Agency and accountability in AI-native teams

Shift from token-maxing to ROI and outcome measurement

Quality framework: 'bad vs. sad'

Why all managers start as ICs at Anthropic

Preventing skill atrophy and fostering team connection

Managing context switching with multiple AI agents

Transformation of PM and data science roles

Importance of dogfooding and using your own product

Unsolved problems in AI-native team operations

The future of engineering jobs and education

Maintaining team culture at scale

Shift from six-month roadmaps to just-in-time monthly planning

Lightning round

Dogfooding

The practice of a company using its own products internally. Fiona emphasizes its importance for leaders and teams to stay connected to user experience, identify issues, and foster product sense.

Latent Demand

Unexpressed or unrecognized needs or desires of customers. Anthropic actively looks for these emerging use cases, even if unintended, to build new products or improve existing ones.

Growth Mindset

A belief that one's abilities and intelligence can be developed through dedication and hard work. This mindset is crucial for adapting to rapid technological changes like AI, encouraging continuous learning and embracing new challenges.

High Agency, High Accountability

A team principle where individuals are empowered with significant freedom and initiative to solve problems ('freedom to cook'), but are also held responsible for the outcomes and hypotheses behind their actions.

Bad vs. Sad Quality Framework

A system for categorizing product quality issues. 'Bad' refers to severe, irrecoverable errors (e.g., crashes), while 'Sad' refers to recoverable pain points or annoyances (e.g., flickering UI). This framework helps teams prioritize and address quality proactively.

Just-In-Time (JIT) Planning

A flexible planning approach that replaces long-term roadmaps with shorter, more adaptive cycles, typically monthly. It focuses on defining immediate priorities and conducting frequent check-ins to adjust as the landscape rapidly changes.

?
How has the engineering role changed over the last 25 years?

The role has transformed from manual coding and hard deadlines (like shipping on CDs) to a world where AI automates much of the code generation, making coding no longer the bottleneck and shifting focus to ambition, verification, and product sense.

?
What does an AI-pilled software team look like in 2026?

Such a team sees roles blurring, with more people across disciplines (designers, PMs) checking in code, high throughput, and a strong focus on verification and proactive quality.

?
How can managers stay on top of the increased code output from AI-native teams?

Managers can use AI tools like Claude routines to automate monitoring feedback channels, track team output, and generate insights or even draft PRs for review, enabling more focused conversations about impact and quality.

?
What qualities are essential for engineers to thrive in an AI-native world?

A strong growth mindset, curiosity, and the willingness to continuously learn and adapt are crucial. Engineers who lean into AI and focus on what's within their control tend to thrive, while resistance leads to frustration.

?
How can small businesses adopt AI tools effectively?

Small businesses can start by identifying specific tasks they dislike or find time-consuming, such as expensing or invoicing, and use AI tools like Claude Co-Work to automate them. Sharing practical use cases among peers helps spread adoption.

?
What is the next frontier for how engineers will operate with AI?

The shift is towards more asynchronous work, leveraging 'routines' that allow AI agents to run tasks and generate prompts autonomously, abstracting away more low-level interactions.

?
How do you measure productivity and ROI of AI tools in today's engineering?

The focus should shift from output metrics (like lines of code or token usage) to outcomes and the problem being solved. Leaders should zoom out, focus on the end goal, and conduct listening tours to understand real impact.

?
How do you balance speed and quality with high velocity AI-generated code?

Prioritize proactive quality measures, such as defining 'bad' (irrecoverable errors) and 'sad' (recoverable pain points) experiences, and investing more in automated monitoring and testing frameworks.

?
Why do managers at Anthropic start as individual contributors (ICs)?

This approach allows new managers to deeply learn the codebase and product, build rapport with the team by understanding their daily work, and gain personal experience with the product before taking on full people management responsibilities.

?
How can engineers prevent skill atrophy when AI writes much of the code?

Engineers should still take time to understand the underlying architecture and dependencies, double-clicking into layers they rely on. Practices like pairwise programming lunches and hackathons also help maintain skills and foster shared learning.

?
What are some of the challenges in the new world of software engineering?

Increased context switching due to multiple AI agents running asynchronously, the potential for loneliness from working primarily with agents, and ensuring foundational understanding of code when AI handles generation are emerging challenges.

?
How should engineering teams approach planning in a rapidly changing AI landscape?

Adopt 'Just-In-Time' (JIT) monthly planning, focusing on short-term priorities with weekly check-ins. This allows for quick adaptation and avoids long-term roadmaps that quickly become obsolete.

1. Embrace AI for Ambition

Use AI tools to overcome perceived technical difficulty, shifting focus from ‘what’s hard’ to ‘how ambitious can you be,’ as AI lifts the ceiling on what’s possible.

2. Lean into Fear with Action

When facing fear or frustration about AI’s impact, ask ‘what can I do about it?’ and ‘what is within my control?’ Taking one small action can counter feelings of helplessness.

3. Cultivate a Growth Mindset

Continuously learn and adapt, recognizing that what made you successful in the past may not serve you in the future. Embrace curiosity and be open to new ways of operating.

4. Share AI Use Cases Widely

To help others adopt AI and prevent a growing divide, share personal examples of how AI tools have made a meaningful difference in your life or work. Start conversations by demonstrating practical benefits.

5. Prioritize Proactive Quality

Implement frameworks like ‘bad vs. sad’ to categorize and address quality issues proactively. Define what constitutes a ‘bad’ (irrecoverable error) and ‘sad’ (recoverable pain point) experience for different product areas, focusing on early detection.

6. Managers Should Stay Hands-On

Leaders should actively use the product their team builds to stay connected to the user experience, understand pain points, and provide meaningful feedback. This builds rapport and ensures the leader remains engaged.

7. Managers Start as ICs

Require new managers to spend time as individual contributors first to deeply learn the codebase and product. This builds rapport with the team and provides a foundational understanding before taking on people management responsibilities.

8. Invest in Automated Verification

As code throughput increases dramatically with AI, shift focus from manual code review to automating checks against well-defined frameworks and specs. This ensures quality at scale and allows human reviewers to focus on deep subject matter expertise.

9. Foster High Agency with Accountability

Empower team members with high agency to solve problems creatively, but pair this with high accountability for the outcomes. Encourage clear hypotheses for what they are trying to solve.

10. Revisit Old Automation Principles

Leverage AI to make previously ’taxing’ but valuable practices, like test-driven development (TDD), more efficient. AI can automate test generation, making these principles more practical to implement.

11. Continuously Question Metrics

Regularly evaluate if existing metrics still serve their intended outcome, as the landscape and product usage can change rapidly. Be prepared to adjust metrics to reflect new realities and avoid blind adherence.

12. Conduct Listening Tours

Leaders should regularly engage in listening tours with senior and junior team members to understand what’s working, what’s not, and how to improve. These conversations can spark ideas and shared learning more effectively than metric dashboards alone.

13. Embrace Asynchronous Work with AI Routines

Leverage AI routines to automate recurring tasks and generate prompts for agents, shifting towards a more asynchronous work style. This allows for higher-level abstraction and delegation of repetitive actions.

14. Combat Loneliness with Pair Programming Lunches

To counter the isolating effect of working extensively with AI agents, implement practices like pairwise programming lunches. This fosters connection, allows team members to learn from diverse workflows, and maintains team cohesion.

15. Explicitly Kill Obsolete Processes

Regularly review and eliminate processes that no longer serve their purpose, especially in a rapidly changing environment. Be open to questioning established norms and adapting planning methods.

16. Adopt Just-In-Time Monthly Planning

Replace long-term roadmaps with lightweight, monthly planning cycles. Define priorities for the upcoming month and conduct weekly check-ins to ensure alignment and adapt quickly to changes.

17. Revisit Past Automation Failures

If an automation attempt with AI previously failed, revisit it with newer models. The exponential rate of improvement means that capabilities that weren’t good enough before might be sufficient now.

18. Foster Openness About Struggles

Create a team culture where managers and team members feel safe to openly discuss what’s not going well, rather than pretending everything is fine. This transparency allows for collaborative problem-solving and prevents issues from festering.

Coding is no longer the bottleneck. It's lifted the ceiling of what anyone is able to do. Everything is now possible in theory. Now it's about how ambitious can you be?

Fiona Fung

For anything that there is a fear, my advice is lean in and ask, what can I do about it? What is within my control?

Fiona Fung

We say with high agency is also high accountability. So it's all about making sure folks have that freedom to cook. But then it's also like, okay, what's the accountability for it? What's a hypothesis of what you're trying to solve?

Fiona Fung

The cave you fear contains the treasure you seek.

Lenny Rachitsky

The fact that that test generation used to just be this tax that I remember having to pay, like, the fact that that's now automated. And you can even revisit all these principles that have been around for a while, but now they actually might be even more efficient just because you have the models that can do more of the work for you.

Fiona Fung

Don't forsake motion for progress.

Fiona Fung

In a world where you can be anything, be kind.

Fiona Fung

Manager's AI-Assisted Team Review

Fiona Fung
  1. Enlist a Claude Code remote session in all team repositories to gain full visibility into everyone's work, Slack channels, and metrics.
  2. Once a month, conduct a shared Claude Code session with the team, sharing your screen.
  3. Review focus areas, shipped products, market performance, and feedback channels together.
  4. Use insights from Claude to identify themes, potential gaps, and areas for future investment, especially regarding quality.

Manager's Daily Feedback Routine

Fiona Fung
  1. Set up an AI routine (e.g., with Claude) to run every morning at a certain time.
  2. Configure the routine to monitor all feedback channels (internal, emails, Slack, social media, partnerships).
  3. The routine should summarize themes from the feedback and generate potential pull requests (PRs) or bug fixes.
  4. Review the AI-generated summary and PRs upon waking, allowing for quick action or further review.

Just-In-Time (JIT) Monthly Planning

Fiona Fung
  1. Define high-level themes for the next six months (e.g., every six months).
  2. For monthly planning, create a lightweight spreadsheet outlining the highest priorities for the upcoming month.
  3. Conduct very quick weekly check-ins to confirm that the monthly priorities are still relevant and aligned with the rapidly changing landscape.
8 times
Average code per quarter increase for Anthropic engineers Compared to 2021-2025
Over $100 billion
Facebook Marketplace annual GMV (Gross Merchandise Volume) As of the time Fiona Fung worked there
Over 25 years
Fiona Fung's engineering experience Started at IBM, then Microsoft, Meta, and Anthropic
Over 500 people
Fiona Fung's org size at Meta/Instagram Overseeing infrastructure, growth, integrity, and safety teams
2006
Estimated Twitter launch year Fiona Fung's recollection
2017
Fiona Fung's last time shipping production software at Meta Before joining Anthropic
8 or 9
Fiona Fung's age when she watched 'Nausicaä Valley of the Wind' Influenced her leadership principles