Anthropic announced Claude Managed Agents last week, a new platform designed to streamline AI agent deployment for enterprises by embedding orchestration logic directly into the AI model layer. The platform allows companies to deploy agents in days rather than weeks or months by handling complex technical requirements including state management, execution graphs, and routing without requiring separate sandboxing or credential management systems.

The new offering represents an architectural shift in how enterprises manage AI agents. Rather than using external orchestration frameworks, companies can now rely on Anthropic's integrated approach where session data is stored in databases managed by Anthropic. This simplification comes with trade-offs, as enterprises cede some control over their AI operations to the model provider, potentially creating vendor lock-in scenarios.

VentureBeat research conducted in early 2025 found Microsoft's Copilot Studio leading enterprise orchestration adoption at 38.6%, followed by OpenAI at 25.7%. Anthropic's tool-use and workflows API grew from 0% to 5.7% adoption between January and February, indicating increasing enterprise interest in the company's native orchestration tools.

Separate research from Databricks addresses a key technical challenge in enterprise AI deployment. The company's study found that multi-step agentic approaches outperformed single-turn retrieval-augmented generation (RAG) systems by over 20% on Stanford's STaRK benchmark when handling queries that combine structured and unstructured data. Even when researchers used stronger foundation models, single-turn systems still underperformed multi-step agents by 21% to 38%.

Databricks' Supervisor Agent demonstrates how multi-step approaches handle complex queries by decomposing questions into parallel SQL and vector search operations, then combining results. This architecture allows agents to process queries like analyzing declining sales data alongside customer review sentiment without requiring data normalization.

The pricing models for these platforms vary significantly. Claude Managed Agents uses a hybrid approach charging $0.08 per hour when agents are active, while Microsoft's Copilot Studio offers capacity-based billing starting at $200 monthly for 25,000 messages. OpenAI's Agents SDK is open-source but charges for underlying API usage at $2.50 per million input tokens and $15 per million output tokens for GPT-4 usage.