Enterprise AI Buyers Put Control Ahead of Platform Loyalty
New VB Pulse data shows enterprises are spreading AI agent orchestration across multiple platforms, with visibility, permissions, and spending controls now shaping buying decisions. For managers, the message is clear: AI workflows need governance and trained human oversight before they need another tool.
Enterprise AI leaders are no longer acting as if one orchestration platform will manage every agent, model, and workflow. A new VB Pulse analysis from VentureBeat suggests that large organizations are deliberately building mixed AI stacks, not simply testing alternatives on the side.
The research, based on responses from 107 enterprises, found that the median enterprise now uses three orchestration tools. That matters for employers because orchestration is the layer that coordinates AI agents, connects them to tools and data, and governs how tasks are executed. In plain terms, it is where a promising assistant becomes part of an operating process.
The practical signal for managers is not that every team should add more platforms. It is that enterprises are prioritizing control, observability, and permissioning as AI moves closer to production work.
Multi-platform AI is becoming the default
According to the survey, 85% of enterprises use two or more orchestration tools, while 64% use three. Only 15% rely on a single orchestration platform.
Microsoft AI Foundry and Copilot Studio appear in 70% of enterprise stacks in the survey. OpenAI's Agents SDK follows closely at 68%, while Anthropic's Claude Platform appears in 47%. Respondents also reported using Google's Enterprise Agent Platform, LangChain and LangGraph, Salesforce Agentforce, Amazon Bedrock, and LlamaIndex. In addition, 22% have built custom orchestration capabilities internally.
The pattern looks intentional. More than half of respondents, 53%, expect their primary AI control plane to be hybrid by the end of 2026. Smaller groups expect a provider-managed service, at 14%, a custom in-house control plane, at 13%, or outside platforms separated from model providers, at 11%.
Many organizations are still in motion. More than two-thirds of respondents expect to change platforms within a year: 15% within three months or sooner, 24% in three to six months, and 28% in six to 12 months. The Claude Agent SDK is under consideration by 43% of builders. About one-third are looking at Google's Enterprise Agent Platform, 31% are considering custom in-house orchestration, and 25% are examining OpenAI options.
Respondents are broadly positive about what they use today, giving current platforms an average overall satisfaction score of 4.17 out of 5. The weaker scores came in ease of implementation, at 3.91, and value for money, at 3.63. For business leaders, those figures are a reminder that platform adoption is not the same as operational maturity.
Spending is shifting toward guardrails
The survey shows enterprise buyers weighing more than model capability when making orchestration decisions. Flexibility was the top factor, cited by 29% of respondents. Security and permissions followed at 17%, production reliability at 15%, and control over agent execution at 15%.
By comparison, only one in 10 respondents named model gravity, meaning tight alignment with a leading base model, as a key buying factor. Ease of development was cited by 8%, total cost of ownership by 4%, and latency and memory performance by 2%.
Budgets reflect the same concern. The largest spending category is agent monitoring and debugging, at 31%, followed by security and permissions enforcement, at 30%. Workflow tooling accounts for 19%. VentureBeat noted that this marks a change from its prior survey wave one month earlier, when workflow tooling led orchestration spending.
The operational priorities are also revealing. Respondents said they are optimizing for task completion reliability, at 30%, multi-step workflow management, at 27%, developer productivity, at 23%, and operational stability, at 13%. Only 7% put end-user experience first.
That does not mean employee experience is unimportant. It suggests many enterprises are still building the underlying machinery before refining how people interact with it. In a human-led AI workplace, that sequence makes sense only if employees are involved in designing the workflow, defining escalation points, and deciding which tasks need human judgment.
The cost and visibility problem is still unresolved
The biggest platform concerns are about oversight. Security and permissioning limits were cited by 37% of respondents, vendor lock-in by 23%, limited visibility and observability by 22%, and lack of flexibility around models and tools by 16%.
Cost control remains a notable gap. One in five enterprises still cannot stop a runaway agent's spending in real time. Among the approaches in use, 30% rely on native platform controls such as budget limits or throttling. Another 25% have built custom gateway plumbing to intercept agents before costs escalate. A further 25% use dynamic routing to send heavier work to lower-cost models. But 21% still depend only on after-the-fact monitoring, such as logs, with no real-time kill switch.
Company size does not appear to solve the issue. Among enterprises with more than 10,000 employees, 18% use only reactive fiscal controls, compared with 23% of smaller organizations.
The survey also suggests that many so-called agents are still early in their development. Only 2% of respondents said 76% to 100% of their systems are advanced and largely autonomous. Another 14% said 51% to 75% are complex, multi-agent pipelines. The largest group, 47%, said 26% to 50% of their systems represent true orchestration. Meanwhile, 35% said only 1% to 25% qualify, and 3% are still deploying chatbots only.
For employers, the conclusion is practical rather than futuristic. Enterprises are not just buying AI agents. They are building governance systems around them. The winners will be teams that can see what the technology is doing, limit what it is allowed to do, and keep skilled people accountable for the workflow around it.
Reported by Hybrion Insights with reference to VentureBeat AI.
hybrion
Better together. Built for what's next.
Stay updated
Get insights on AI, teams, and the future of work.
