Anthropic is shifting more decisively toward business customers as its Claude models gain traction. The company is preparing a multibillion-dollar funding round that could push its valuation toward $950 billion — approaching the $854 billion valuation assigned to OpenAI in its March funding round.
A recent report says Anthropic may have overtaken OpenAI among enterprise customers: the company has quadrupled its market share since May 2025. Cat Wu, head of product development for Anthropic’s Claude Code and Cowork, who joined the company in August 2024, is credited as a key driver of that growth.
Claude's evolution: from chatbot to coding and productivity platform
Wu helped guide Claude through a critical transformation from a simple informational chatbot into a tool for coding, programming assistance, and broader work workflows. She frequently collaborates with Boris Cherny, a central member of Anthropic’s technical team and the creator of Claude Code; internally the pair is often nicknamed the company’s “Batman and Robin.”
Speaking at the Code with Claude conference, Wu described how the team frames product strategy around the expectation of exponential technological improvement rather than closely tracking competitors. She argued that over-focusing on rivals can slow execution, and that staying at the technological forefront requires concentrating on rapid, continuous innovation.
Wu said the pace of model development still looks steady, and Anthropic expects to continue releasing models and features regularly. At the same time, she emphasized cautious rollout practices: some capabilities are introduced in limited ways for safety and control, as illustrated by the Glasswing program.
Glasswing and Mythos: restricted access for safety reasons
Anthropic launched the Glasswing initiative in April, giving a narrow partner consortium — including Amazon, Apple, CrowdStrike and Microsoft — access to its new cybersecurity model, Mythos. Unlike other Anthropic models, Mythos will not be publicly released: the company says it is concerned that a model designed to search for software vulnerabilities could be too powerful and potentially weaponized by malicious actors.
Agents and the future of work
Wu has previously said that much of the future of work will involve people managing fleets of AI agents. She noted that effective oversight requires domain expertise: it is difficult to direct agents well if you do not understand the work yourself. Managers will need a new skill set akin to debugging an agent — diagnosing why it made an error, whether it misunderstood instructions, or whether the prompt was insufficiently precise.
While that raises questions about workforce changes — for example, whether fewer interns or junior staff will be needed for repetitive tasks — Wu framed the ideal outcome as augmentative: agents should take over monotonous parts of jobs so people can focus on higher-value, creative activities.
Near-term focus: proactivity
Wu believes the next major advance over the coming months will be proactivity. Until recently, AI systems operated mainly in a synchronous, request-driven mode. Users are now automating routines such as customer-support responses. The next phase will be Claude understanding what a user is working on and autonomously setting up certain automations for them, while maintaining safety and usability safeguards.
Why this matters
Anthropic’s rising enterprise adoption and the rapid growth of Claude suggest the large language model landscape may support multiple dominant players. The company’s high valuation target, its controlled partner rollouts for sensitive models, and its push toward proactive, autonomous agents are indicators of how AI deployment in business could evolve in the months ahead.
(Interview excerpts are based on the edited version published by techcrunch.com.)
