Key Takeaways
- AI-driven productivity can lead to engineer isolation. Fiona Fung, who leads Claude Code at Anthropic, observed that her team started feeling lonely due to extensive solo work with agents, prompting new social initiatives.
- Leaders must "dogfood" the product to stay connected. Fung noted she hadn't shipped production software at Meta since 2017, highlighting the need for hands-on engagement to understand the current developer experience with tools like Claude Code.
- Team culture and agility are critical counterweights to AI's impact. Fung emphasizes a "one team mentality" and flexible structures, like adopting asynchronous work and increasing engineer agency, to maintain cohesion.
- Traditional long-term planning is dead for fast-moving AI teams. Anthropic shifted from 6-month plans to "JIT planning" (just-in-time planning), aligning monthly on a simple spreadsheet rather than dense documentation.
- Agility hinges on actively removing obsolete workflows, formalized by the “Explicit Permission to Kill Processes That No Longer Serve Us” framework.
The "Explicit Permission to Kill Processes That No Longer Serve Us"
Regular Review: Regularly assess all team processes, rituals, and documentation.
Question Purpose: For each process, ask: “Is it still serving its purpose?” Is it highly noisy, expensive, or manual without clear value?
Empower Discontinuation: Grant explicit permission to team members to propose and (with alignment) eliminate processes that are no longer effective or relevant given the fast-changing landscape.
When This Works (and When It Doesn't)
Fung highlights this framework as essential for maintaining agility and preventing "process bloat" in rapidly evolving environments. This approach shines in high-velocity startups, R&D labs, or any team where market conditions, technology, or internal structures change often. It ensures team energy focuses on high-value work, not legacy routines.
However, this framework falls short in highly regulated industries or environments demanding strict process adherence for safety, compliance, or quality control. If your work involves ISO certifications, government contracts, or medical devices, killing a process without thorough impact assessment might invite disaster. Use it where flexibility trumps rigidity, but exercise caution where consistency is paramount.
What to Do With This
Your weekly all-hands meeting used to be a buzzing hub of energy. Now it's a 90-minute status report where everyone zones out. You, a 27-year-old founder, feel its dead weight. Use Fung's framework to tackle it this week.
1. Regular Review: List out the meeting's agenda, attendees, and stated purpose. Look at the last four meeting notes.
2. Question Purpose: Ask yourself and a few trusted team members: "Is this meeting still serving its purpose of aligning the team and making decisions, or is it just a noisy, expensive sync-up?" Fiona Fung suggests this direct questioning to expose routines that have lost their way. Perhaps it's become a broadcast when it should be a discussion.
3. Empower Discontinuation: At your next leadership sync, say, "I'm giving us explicit permission to kill processes that no longer serve us. Our weekly all-hands feels like one of them. I propose we scrap it next week. Instead, we'll try a 30-minute 'Decision Point' meeting on Tuesday and move status updates to an async Slack thread." See what happens. The key, Fung shared, is actively granting that permission, making it safe to challenge the status quo. Anthropic's team shifted from 6-month plans to lightweight monthly spreadsheets for "JIT planning" because long cycles no longer served them; your all-hands might be the same. Embracing this means giving engineers more agency and trusting their input, a lesson Anthropic applies to counter potential AI-induced loneliness by fostering connection even in new ways, like "pairwise programming lunches." Don't just prune; be ready to cut.