Enterprises standardize agent orchestration before agents mature

VentureBeat Pulse Research surveyed 101 enterprises and found agent orchestration concentrating around model-provider platforms, with Anthropic’s Claude named the primary platform by 40% of respondents. The finding matters because enterprises are spending and planning around agent control layers even though 71% said no more than a quarter of their deployed “agents” are true multi-step workflows.

The survey, drawn from a single June 2026 wave of organizations with at least 100 employees, should be read as directional rather than a spend-weighted market measure. VentureBeat said the respondents were self-selected and included product and program managers, CIOs, CTOs, CISOs, consultants, and data, AI and engineering leaders. Technology and software companies made up 44% of the sample, followed by financial services at 17% and healthcare and life sciences at 8%.

Model providers are winning the first deployment phase

Among primary orchestration platforms, Anthropic led by a wide margin at 40%. Microsoft followed at 18%, and OpenAI at 13%. Google and Amazon were also included among the major model platforms, which together accounted for about four-fifths of primary deployments in the survey. Open frameworks such as LangChain and LangGraph, along with custom internal builds, remained in single digits.

VentureBeat attributed that concentration to what it called “model gravity”: enterprises are choosing orchestration environments tied closely to the models they want to use. When asked what drove platform selection, 21% pointed to native alignment with a leading base model. Flexibility across models and tools and ease of development each drew 17%.

The platform concentration does not mean buyers are settled. VentureBeat found that 68% plan to adopt a new, additional or replacement orchestration platform within 12 months. Among those considering changes, the largest group, 29% of all respondents, had not yet picked a shortlist. OpenAI led named candidates at 16%, followed by LangChain or LangGraph at 12% and Anthropic at 7%.

The deployed agent base is less advanced than the strategy

Enterprises said they judge orchestration mainly on whether it can finish multi-step work. Task completion reliability was the top success metric at 32%, while multi-step workflow management followed at 28%. Developer productivity, often prominent in tool discussions, was selected by 17%.

The gap is that many deployments are not doing the work those metrics describe. VentureBeat reported that 71% of enterprises said a quarter or fewer of their deployed agents are genuine multi-step orchestrated workflows rather than single-prompt chatbot wrappers. Only 10% said more than half of their agents had reached that level.

Company size appears to matter. VentureBeat said 77% of smaller enterprises reported that a quarter or fewer of their agents do multi-step work, compared with 62% of larger enterprises. The report characterized those figures as directional.

Enterprises want hybrid control and better cost limits

By the end of 2026, 51% of respondents expect agent control to sit in a hybrid setup combining provider-native tooling with external orchestration. Only 6% expect to rely fully on a provider-managed service. Asked about the main risk of putting control inside a model-provider platform, 35% cited vendor lock-in, ahead of security and permissioning limits at 28%.

Budgets are following the same pattern. Agent workflow tooling was the top area expected to see increased investment, named by 34% of respondents. Security and permissions enforcement followed at 25%, with scaling infrastructure at 20%.

Cost controls are less mature. VentureBeat found that 27% of enterprises have no real-time, programmatic way to stop a runaway agent before the bill arrives. Another 32% rely on native caps and throttles from their primary platform. Smaller enterprises were more likely to have reactive controls only, at 34%, compared with 20% of larger enterprises.

The survey points to a market where orchestration architecture is being built ahead of the agent portfolios it is supposed to manage. Providers have the early deployment advantage, but buyer intent, lock-in concerns and weak fiscal controls suggest the enterprise agent stack remains unsettled.

This story draws on original reporting from VentureBeat.