Quick Facts
- 69% of enterprises using OpenAI's Agents SDK or Responses API name it their primary orchestration platform, versus 38% for Anthropic's Claude Platform.
- 85% of enterprises run two or more agent orchestration platforms, with a mean of 3.1 platforms per organization.
- Anthropic leads the forward pipeline among enterprises planning a platform change, with 43% of those 72 organizations considering it next.
Enterprises are installing AI agent platforms at high rates. What they do after installation tells a different story depending on the vendor.
A new VentureBeat VB Pulse survey from August 2026, drawing 169 qualified responses from organizations with 100 or more employees, found that 52 of the 75 enterprises running OpenAI's Agents SDK or Responses API — 69% — named it their primary orchestration platform. Of the 45 enterprises running Anthropic's Claude Platform and Agent Skills, only 17, or 38%, did the same. That 38% is the lowest conversion rate of any platform in the study.
The distinction matters. A primary platform controls which agents run, what tools and data they can access, and what happens when a step fails. Installing a platform and building on it are two different decisions.
Where the Platforms Stand
On primary platform share, OpenAI leads the August wave. Fifty-three of 162 enterprises that named a primary platform chose OpenAI, representing 33%. Google's Enterprise Agent Platform ranked second at 24%, and Anthropic came in third at 11%.
Stack presence tells a different story. Microsoft AI Foundry and Copilot Studio appear in 70% of enterprise agent stacks. OpenAI's Agents SDK shows up in 68%, and Anthropic's Claude Platform in 47%. The market is not choosing one winner — it is stacking several platforms at once. The survey found 85% of enterprises run two or more orchestration platforms, 64% run three or more, and the mean sits at 3.1 platforms per organization.
Anthropic's Pipeline Advantage
Anthropic's weakest number is its convert rate. Its strongest number is what comes next. Among 72 enterprises in motion for a new platform, 43% named Anthropic as their target — well ahead of Google at 31%, in-house builds at 31%, OpenAI at 25%, LangChain/LangGraph at 17%, and Microsoft at 17%.
Anthropic also had 36 organizations considering adoption within a year, compared with its 45 current users. OpenAI had 41 prospective adopters against 75 current users. Both ratios suggest continued demand, but Anthropic's pipeline is proportionally larger relative to its installed base.
Market Signals Outside the Survey
Spending data from other trackers adds more color. The Ramp AI Index for August 2026 shows 43.8% of US businesses paying for Anthropic subscriptions or tokens, compared with 39.8% for OpenAI. Anthropic gained 0.34 percentage points month over month; OpenAI gained 0.09.
On OpenRouter, however, the picture shifted. In the week of September 7, 2026, OpenAI models captured just over half of the combined dollar spend between the two labs — the first time OpenAI had led that metric in more than two and a half years. The VentureBeat survey's authors note these spending figures cannot distinguish between a platform an enterprise is testing and one it is building on.
What Executives Are Watching
Sixty percent of survey respondents said they plan to adopt, add, or replace an agent orchestration platform within a year, including 47 organizations planning a move within the quarter. That volume of planned change means the current rankings are not stable.
Vendor inflexibility across models and tools ranked as the top concern at 28% of respondents. Vendor lock-in ranked fourth at 16%. Asked where the primary control plane for enterprise agents will live by the end of 2026, 33% said a hybrid model and 26% said a provider-managed agent service.
One additional finding cuts through the adoption headlines: 47% of respondents said only 26% to 50% of their so-called agents are genuinely orchestrated. Thirty-seven percent put that share at 25% or below. Most companies are still running chatbots, not agents.
This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.