TechCrunch: Anthropic's head of product on AI that acts before you ask

On May 13, 2026, TechCrunch published an interview with Cat Wu, Anthropic’s head of product for Claude Code and Cowork, conducted at the Code with Claude conference. Wu discussed Anthropic’s product strategy, changes in how enterprises are adopting AI, and what she described as the next major design challenge: proactivity.

Anthropic’s position

Anthropic has seen substantial growth in enterprise adoption over the past year, with business customers increasingly choosing Claude over competing products. Wu attributed this partly to a deliberate focus on capability development rather than reactive competitive positioning — “staying on the exponential” in her phrasing. The company released at least six models in 2025 and is tracking a similar pace in 2026.

The proactivity shift

The most consequential part of the interview concerns what Wu identifies as the next phase of AI product design. Current AI products are synchronous: a user makes a request and the AI responds. Wu’s view is that the meaningful next step is AI that understands a user’s ongoing work context and takes actions or sets up automations before the user formulates a request. Her example: Claude noticing patterns in how a user works and configuring relevant automations without being asked.

This creates a distinct design challenge for product managers. Proactive AI requires products to infer user intent from context rather than explicit prompts, which raises questions about transparency, consent, and how to present AI-initiated actions in a way users can understand and reverse. These are product specification problems, not only ethical ones.

On workforce impact

Wu’s comments on work are precise: she acknowledged that AI will automate routine tasks, using email responses as an example, but argued that the goal should be redirecting freed capacity toward creative work rather than reducing headcount. For product teams deciding how to position AI-powered features internally or externally, that distinction matters both for messaging and for how success is measured.