VentureBeat Pulse Research surveyed 101 enterprise respondents in June 2026 (organizations with 100+ employees) about agent orchestration. The data show rapid consolidation onto major model‑provider platforms—Anthropic’s Claude is the primary orchestration platform for 40%—but a large gap between strategic ambition and deployed reality: most deployed "agents" remain single‑prompt chatbot wrappers rather than genuinely orchestrated multi‑step workflows.
Key findings
- Platform concentration: Anthropic (Claude) leads with 40%, followed by Microsoft at 18% and OpenAI at 13%; Anthropic, Microsoft, OpenAI, Google and Amazon together account for roughly 80% of primary deployments. LangChain/LangGraph and custom in‑house builds are single‑digit shares. 3% report no orchestration.
- Drivers of platform choice: “Model gravity” (native alignment with a state‑of‑the‑art base model) is cited by 21% as the primary influence, followed by multi‑model/tool flexibility and ease of development (17% each), security and permissions (14%), and total cost of ownership (11%).
- Success metrics: enterprises judge orchestration by reliable, multi‑step execution—task completion reliability (32%) and multi‑step workflow management (28%) are the top metrics. Developer productivity (17%) and end‑user experience (9%) are less frequently prioritized.
- The chatbot trap: 71% of respondents say a quarter or fewer of their deployed "agents" are true multi‑step orchestrated workflows; only 10% report more than half of their agents are genuinely orchestrated.
Architecture and strategic moves
- Control plane expectations: a majority (51%) expect a hybrid control plane (provider‑native plus external orchestration) by the end of 2026. Only 6% expect to hand primary control entirely to a provider‑managed service. Overall, architectures that keep control at least partly outside the provider sum to 88%.
- Lock‑in concern: vendor lock‑in is the top risk if control sits inside a model provider (35%), followed by security/permission limitations (28%) and inflexibility across models/tools (21%).
- Planned changes in the next 12 months: building in‑house control planes (25%), standardizing on a single framework (24%), and moving agents from sandbox to production (23%) are nearly tied at the top, indicating a shift from experimentation to operational consolidation.
- Investment priorities: spending is highest on agent workflow tooling (34%), then security and permissions enforcement (25%) and scaling infrastructure (20%). Monitoring and debugging receive smaller shares (11% each for monitoring and flat budgets).
Fiscal control and runtime risk
- Real‑time cost control is uncommon: 27% of enterprises have no programmatic real‑time way to stop a runaway agent before the bill arrives and learn of excessive token consumption only from logs after the fact. 32% rely solely on the provider’s native caps and throttles. Custom gateways (23%) and cross‑model routing for cost arbitrage (19%) are minority approaches.
- Company size matters: about 34% of enterprises under 2,500 employees report only reactive spend control versus roughly 20% of larger enterprises, aligning with the distribution of orchestration maturity.
Methodology
VentureBeat fielded this Pulse Research wave in June 2026. The instrument targeted organizations with 100 or more employees and returned 101 responses. Respondents include product and program managers, CIO/CTO/CISO, consultants/advisors, and directors/VPs of data, AI, and engineering. Industry distribution: Technology/Software 44%, Financial Services 17%, Healthcare/Life Sciences 8%, and others. The sample is self‑selected and cross‑sectional; results are directional rather than a probability‑based market measure.
Conclusion
Enterprises are consolidating orchestration around major model‑provider platforms—Anthropic’s Claude leads—but the deployed estate is largely not yet doing the multi‑step work orchestration is meant to deliver. Organizations are funding workflow tooling and permissions, planning hybrid control planes to hedge against vendor lock‑in, and moving agents toward production. However, 71% report that a quarter or fewer of their agents are genuinely orchestrated, only 10% exceed the halfway mark, and over a quarter cannot stop runaway agent spending in real time. In short: the orchestration layer is being built before most agents actually need it.
Questions to watch
- Will the gap between orchestration ambition and deployed multi‑step workflows close quickly?
- How rapidly will enterprises adopt programmatic real‑time fiscal controls for token consumption?
- Will preferences for hybrid control persist as provider features and pricing evolve?
Subsequent Pulse waves will show whether the "chatbot trap" is transient or a more persistent characteristic of enterprise agent deployments.
