One in Five Enterprises Can't Stop a Runaway AI Agent's Spending in Real Time
Taryn Plumb
12:57 pm, PT, August 20, 2026
Enterprise AI teams have moved away from relying on a single orchestration platform. According to VB Pulse data, the median enterprise now runs three platforms simultaneously, not out of necessity, but due to a lack of trust in any one vendor's security and permissioning capabilities.
These findings come from an ongoing analysis of enterprise AI deployments, focusing on:
- Platforms in use
- Decision-making factors
- Priorities
- AI expectations
- Cost control
- The extent to which AI agents exhibit true agency versus being limited chatbots.
VB Intelligence gathers insights directly from software and machine learning engineers, product and program managers, and data/AI/analytics VPs and directors.
Concerns Around Retaining Visibility and Control
Agentic orchestration has become plural across 107 enterprises surveyed:
- 85% use two or more orchestration tools
- 64% employ three different platforms
- Only 15% operate on a single orchestration platform.
The most popular tools include:
- Microsoft AI Foundry/Copilot Studio (70%)
- OpenAI’s Agents SDK (68%)
- Anthropic’s Claude Platform (47%)
Other notable mentions are Google’s Enterprise Agent Platform, LangChain/LangGraph, Salesforce Agentforce, Amazon Bedrock, and LlamaIndex. Additionally, 22% of builders run custom in-house orchestration solutions.
This trend towards hybridization is expected to grow:
- 53% predict the primary control plane will be hybrid by year's end.
- 14% plan to use a provider-managed service
- 13% aim for a custom in-house control plane
- 11% rely on external platforms abstracted from model providers.
More than two-thirds ( 68%) intend to switch platforms within the year, with:
- 15% planning a change in the next three months (or sooner)
- 24% considering a shift in 3 to 6 months
- 28% looking to make a move in 6 to 12 months.
The Claude Agent SDK from Anthropic is a top choice for exploration, with 43% of builders considering it. Others under consideration include Google’s Enterprise Agent Platform ( 31%), custom in-house orchestration ( 31%), and OpenAI’s options ( 25%).
Enterprises are avoiding vendor lock-in, opting for a hybrid approach where multiple platforms, models, and agents work together across a shared control plane.
Overall, respondents rate existing platforms highly (4.17 out of 5) but express lower satisfaction with ease of implementation (3.91 out of 5) and value for money (3.63 out of 5). These ratings will be interesting to watch as orchestration platforms and AI roadmaps evolve.
Where Enterprises Are Investing Their Money
Enterprise buying decisions are guided by:
- Flexibility (29%)
- Security and permissions (17%)
- Production reliability (15%)
- Control over agent execution (15%).