That’s proven harder than expected. Enterprises experimenting with agents have run into a familiar set of challenges, from maintaining context across sessions to coordinating multistep workflows, integrating with internal systems like CRMs and databases and staying within strict security and compliance guardrails.
Most companies have not built that infrastructure themselves Instead, they have turned to a growing category of startups offering orchestration layers, workflow engines, and agent management tools, a category Claude Managed Agents is now taking on more directly.
The offering is now in public beta. Early customers Notion, Rakuten and Asana have used the platform to deploy agents across functions including project management, HR, finance and software development.
From Models to Execution
The shift reflects a broader transition in how AI systems are deployed internally.
Early adoption centered on using large language models for chat interfaces, copilots and isolated automation tasks. The next phase is execution, where AI systems are expected to complete end-to-end workflows with minimal human input.
That shift also introduces complexity. A single task, such as processing an insurance claim or handling a customer support escalation, can require multiple steps, decision points and system integrations.
Agents must retrieve and update data, call external tools and adapt based on intermediate results. Maintaining consistency across those steps, especially over long-running tasks, has been one of the hardest problems to solve.
Startups Under Pressure
Startups have emerged to address this gap. Many offer frameworks that allow developers to define workflows, manage memory and integrate models with enterprise systems. Others focus on reliability, adding guardrails, monitoring and retry mechanisms to reduce failure rates.
Anthropic’s approach is to bundle those capabilities into its own platform. Claude Managed Agents lets enterprises define tasks, connect tools and deploy agents without building a separate orchestration layer.
The result: less need to stitch together multiple vendors or build custom infrastructure, while shifting more of the application logic closer to the model itself.
The agent infrastructure market has been one of the busiest corners of enterprise AI over the past two years. Agentic AI startups attracted $2.8 billion in venture funding in the first half of 2025 alone.
For example, Sierra, which builds customer service AI agents and was co-founded by former Salesforce co-CEO Bret Taylor and ex-Google VP Clay Bavor, raised $350 million at a $10 billion valuation in September 2025. The company reached $100 million in annual recurring revenue in under two years.
As recently as last week, Sycamore raised a $65 million seed round led by Coatue and Lightspeed to build what its founder described as a full agentic orchestration layer for enterprise deployments.
For enterprises, the calculus shifts from assembling multiple vendors to buying the capability in one place. For startups selling the missing pieces, that’s a harder position to defend.
Stickier Than an API
The commercial logic is straightforward. Selling model access generates revenue, but it does not make Anthropic hard to replace. A managed platform does. Once a company’s agents run on Anthropic’s infrastructure, the workflows and operational setup become embedded in how the business runs, raising the cost of switching, according to Startup Fortune.
Anthropic crossed a $30 billion annualized revenue run rate this month, up from roughly $9 billion at year-end 2025, driven largely by enterprise demand as reported by PYMNTS.
Claude Managed Agents extends that relationship further up the stack, from model provider to execution environment.
It’s a familiar pattern in enterprise technology. Cloud providers spent a decade absorbing functions like database management, deployment pipelines and monitoring that had been handled by separate vendors. The companies that built those middle-layer tools either differentiated or got folded into the platform. AI infrastructure is following the same arc.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
