- Microsoft’s Magentic Marketplace reveals AI agents’ inability to act independently
- Client-side agents were easily influenced by sales agents during simulated transactions
- AI agents slow down significantly when faced with too many choices
A new study from Microsoft has raised questions about the current suitability of AI agents operating without full human supervision.
The company recently built a synthetic environment, the “Magentic Marketplace,” designed to observe the performance of AI agents in unsupervised situations.
The project took the form of a fully simulated e-commerce platform that allowed researchers to study how AI agents behave as customers and businesses – with predictable outcomes possible.
Testing the limits of current AI models
The project included 100 client-side agents interacting with 300 enterprise-side agents, providing the team with a controlled environment to test the agents’ decision-making and negotiation skills.
The market’s source code is open source; therefore, other researchers can adopt it to reproduce experiments or explore new variants.
Ece Kamar, vice president and general manager of the AI Frontiers Lab at Microsoft Research, noted that this research is essential to understanding how AI agents collaborate and make decisions.
Initial testing used a mix of leading models including GPT-4o, GPT-5 and Gemini-2.5-Flash.
The results were not entirely unexpected, as several models had weaknesses.
Customer agents could easily be influenced by sales agents in product selection, thereby revealing potential vulnerabilities when agents interact in competitive environments.
Agents’ effectiveness declined sharply when they were faced with too many options, overwhelming their attention span and leading to slower or less accurate decisions.
AI agents also struggled when asked to work toward common goals, as the models were often unclear which agent should take which role, reducing their effectiveness in common tasks.
However, their performance only improved when step-by-step instructions were provided.
“We can instruct the models, like we can tell them, step by step. But if we’re inherently testing their collaboration capabilities, I would expect those models to have those capabilities by default,” Kamar noted.
The results show that AI tools still need significant human support to operate effectively in multi-agent environments.
Often presented as capable of making independent decisions and collaborating, the results show that the behavior of unsupervised agents remains unreliable. Humans must therefore improve coordination mechanisms and add safeguards against AI manipulation.
Microsoft’s simulation shows that AI agents are far from operating independently in competitive or collaborative scenarios and may never achieve full autonomy.
Follow TechRadar on Google News And add us as your favorite source to get our news, reviews and expert opinions in your feeds. Make sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp Also.
