Claude from Anthropic Dominates Agentic Orchestration

⚡

Key Takeaways

1Claude by Anthropic is the preferred platform for 40% of companies, surpassing Microsoft and OpenAI.

271% of companies admit that their agents are merely chatbots, not orchestrated workflows.

351% of companies plan to implement hybrid control by 2026 to avoid vendor lock-in.

💡Why it matters — Companies need to overcome orchestration challenges to maximize the efficiency of their AI agents while avoiding reliance on a single vendor.

Claude from Anthropic Leads Agentic Orchestration

A recent study conducted among 101 companies reveals that the orchestration of AI agents in business primarily revolves around model provider platforms. Among these, Claude from Anthropic stands out as the preferred choice. This platform is selected for the robustness of its underlying model and its ability to reliably execute complex tasks. However, despite these high ambitions, the reality is that most of the "agents" in place are still just chatbot interfaces. Companies are opting for a hybrid control approach to avoid being locked in by a single provider, and real-time tracking of token consumption remains rare.

The VentureBeat Pulse study examines the orchestration of agents in business, analyzing the platforms used, selection criteria, optimized aspects, agent control structures, and the true orchestration of deployed agents. It also looks at how companies manage the costs associated with these agents.

A Gap Between Ambition and Reality

The main finding of this survey is the significant gap between companies' orchestration ambitions and the reality of their implementation. Companies are turning en masse to the leading model platforms, with Claude from Anthropic at the forefront, chosen by 40% of companies, more than double that of its direct competitors, Microsoft (18%) and OpenAI (13%). This choice is driven by alignment with an advanced base model (21%) and reliable task execution (32%), as well as multi-step workflow management (28%). However, 71% of companies acknowledge that only a quarter or less of their deployed "agents" are true orchestrated workflows, and only 10% have reached a more advanced level of orchestration. The orchestration layer is often implemented even before the orchestrated portfolio is fully operational.

Towards a Hybrid Architecture

This gap between ambition and reality directly influences the architecture that companies adopt. By the end of 2026, a majority of them (51%) plan to implement a hybrid control plan, combining native provider elements with external orchestration. Only 6% of companies consider fully entrusting control to a provider-managed service, primarily due to fears of vendor lock-in (35%). Investments reflect this trend, with a priority given to agent workflow management tools (34%), followed by security and permission enforcement (25%). Real-time fiscal control remains a challenge, with more than a quarter of companies (27%) lacking a means to stop an uncontrolled agent before the bill is issued.

Study Methodology

This survey was conducted by VentureBeat as part of its Pulse research series, focusing on the orchestration of agents in business. Responses were collected from organizations with 100 or more employees (n=101), during a single wave in June 2026. The sample is evenly distributed among different company sizes: 100–499 employees, 2,500–9,999, and 50,000+ (21% each), with 10,000–49,999 and 500–2,499 (19% each). Respondents include senior decision-makers such as product and program managers (15%), CIOs/CTOs/CISOs (13%), consultants and advisors (13%), as well as a range of directors and VPs in data, AI, and engineering, with an "Other" category at 18%. Regarding purchasing, 81% of respondents are recommenders, influencers, or final decision-makers for AI solutions (66% recommender/influencer, 15% final decision-maker). The Technology/Software sector is the most represented with 44%, followed by Financial Services (17%) and Life Sciences/Health (8%).

With 101 respondents, the sample is deemed robust enough to provide reliable conclusions, although it is self-selected and does not constitute a probabilistic sample.

Conclusion

The results of this study show that the orchestration of agents in business is still in its early stages, with an urgent need to improve reliability and workflow management while avoiding vendor lock-in. Companies are seeking to evolve towards more integrated and controlled orchestration, while investing in tools that will enable them to better manage their AI operations.