Why Anthropic says long product roadmaps create dangerous delays

This article summarizes an episode of Lenny’s Podcast’s video series featuring Fiona Fung, product and engineering leader at Anthropic.

Photo credit: Shutterstock

The act of typing code is no longer the primary bottleneck in software development.

Fiona Fung, product and engineering leader for Claude Code and Cowork at Anthropic, contends that successful products require a new approach as AI accelerates production.

Teams must shift their focus toward defining sharper goals, verifying outputs, and restructuring human collaboration.

Imagination replaces coding as the main limit

Software development previously relied on the sheer hours engineers could dedicate to a project. Workers now produce code nearly instantly.

Success hinges on identifying the right problems and understanding user needs, allowing employees to tackle tasks beyond their original training.

“It’s lifted the ceiling of what anyone is able to do,” Fung observes, detailing how a team member lacking mobile experience used the tool to build a complete Android feature.

Business leaders face pressure to update hiring and organizational models, passing tasks between specialized departments only causes delays.

While anyone can attempt unfamiliar tasks, companies still require domain experts to catch errors that might compromise security, break systems, or damage customer trust.

The most effective staff combine creative imagination with market awareness and enough technical knowledge to judge when AI provides useful answers, obscures risks, or misses crucial details.

Freedom requires clear expectations

Unleashing this creative imagination grants employees operational freedom, yet unchecked autonomy often results in wasted effort and confusion.

Workers using AI must take independent action while proving their concepts work, handling everything from anticipating user needs to measuring sales. Leaders must demand clear accountability for results.

“With high agency comes with high accountability,” Fung says, emphasizing that while her team has autonomy to solve problems creatively, members must state their hypotheses and own the outcomes.

Checking work replaces writing code

Managers direct machines instead of people