Why do companies use multiple platforms for AI orchestration?

Businesses avoid vendor lock-in and seek flexibility. They also distrust third-party platforms' security and control capabilities, preferring hybrid solutions.

What are the main challenges with AI agents?

Key issues include token spending control, limited agent activity visibility, and security gaps. Many companies lack effective mechanisms for real-time expense management.

← All news

Artificial intelligence

20% of Large Companies Struggle to Control Unchecked AI Agent Spending

August 21, 2026

Photo: images.ctfassets.net

Quick answer

Enterprises are rapidly adopting AI agents, but 20% cannot stop their uncontrolled token spending in real time.

Large enterprises are rapidly adopting AI agents, but face significant challenges in cost management. According to VentureBeat research, 20% of companies cannot halt uncontrolled token spending in real time, creating financial risks. Most organizations use multiple orchestration platforms to avoid vendor lock-in and maintain process control.

On average, enterprises operate with three platforms simultaneously. The most popular solutions are Microsoft AI Foundry/Copilot Studio (70% of companies), OpenAI Agents SDK (68%), and Anthropic Claude Platform (47%). Meanwhile, 22% of respondents develop custom orchestration solutions to enhance security and cost control. Key platform selection priorities include flexibility (29%), security (17%), and task execution reliability (15%).

Despite high satisfaction with current platforms (average rating: 4.17/5), companies report challenges in implementation and ROI. Major concerns include limited agent activity visibility (22%) and vendor dependency risks (23%). For cost control, 30% of businesses rely on built-in platform limits, while 25% have developed custom intermediary solutions to intercept 'runaway' agents.

Only 2% of enterprises report that 76–100% of their systems are fully autonomous AI agents. Most (47%) state that only 26–50% of their solutions meet this level. Others use basic assistants or advanced chatbots. Experts predict a full transition to multi-step autonomous agents will take several more years.

Common questions

Why do companies use multiple platforms for AI orchestration?: Businesses avoid vendor lock-in and seek flexibility. They also distrust third-party platforms' security and control capabilities, preferring hybrid solutions.
Which AI agent platforms are most popular?: Leading platforms include Microsoft AI Foundry/Copilot Studio (70% of companies), OpenAI Agents SDK (68%), and Anthropic Claude Platform (47%). Google, Salesforce, Amazon, and LangChain solutions are also widely used.
What are the main challenges with AI agents?: Key issues include token spending control, limited agent activity visibility, and security gaps. Many companies lack effective mechanisms for real-time expense management.

Dzen feed: /feed/dzen.xml · RSS: /feed.xml