Enterprise AI teams are running multiple orchestration platforms simultaneously out of distrust, not diversity preference. The median enterprise now operates three separate platforms at once, a defensive strategy revealing deep skepticism about vendor capability in cost control, security, and permissioning.
The finding comes from VB Pulse data on agentic orchestration adoption. One in five enterprises cannot stop a runaway AI agent's spending in real time. This gap in cost governance represents both a technical failure and a business risk as autonomous agents scale across organizations.
Microsoft holds the largest share of primary orchestration usage today, though enterprises are hedging their bets across multiple platforms. Anthropic leads significantly in consideration for next-generation deployments, suggesting enterprises plan to diversify further rather than consolidate.
The multi-platform strategy reflects three distinct pain points. First, vendor lock-in concerns remain real enough to justify operational overhead. Enterprises worry that relying on a single orchestration vendor creates switching costs and reduces negotiating power as agentic AI matures. Second, security and permissioning capabilities across vendors remain immature. Enterprises lack confidence that any single platform can enforce their internal security policies, data governance requirements, and access controls. Third, cost metering and spending limits lack sophistication. The fact that 20 percent of enterprises cannot stop autonomous agents from incurring unexpected costs suggests orchestration platforms lack real-time spending oversight features that traditional cloud infrastructure has provided for years.
This governance gap creates real financial risk. Autonomous agents operating across APIs, cloud services, and third-party systems can accumulate costs faster than human-managed workloads. Without hard spending limits enforced at the orchestration layer, a misconfigured agent or prompt injection attack could trigger thousands in unauthorized expenses before detection.
The multi-platform approach solves the vendor trust problem while creating new operational complexity. Running three orchestration platforms means managing three separate control planes, three different API integrations, three sets of security policies, and three cost-tracking systems. Enterprises are choosing complexity over the alternative: betting their agent infrastructure on a single vendor's governance tooling.
This decision reflects where the agentic AI market stands in mid-2024. The technology moves faster than vendor capabilities in security, compliance, and cost control. No orchestration platform has yet proven it can handle enterprise-grade agent governance at scale. Microsoft's lead in current usage stems from customer existing relationships and cloud integration, not superior agent orchestration. Anthropic's strength in next-consideration signals that enterprises plan to test new players as the market evolves.
The path forward requires vendors to prioritize governance over feature velocity. Real-time spending controls, granular permissioning, audit trails, and drift detection matter more than additional model parameters or faster inference. Enterprises will consolidate back to fewer platforms once any vendor solves the governance problem comprehensively. Until then, expect the three-platform median to hold steady or grow.