Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Join a reliable event by enterprise leaders in about two decades. VB Transform, Real Enterprise AI strategy brings together people who build. Learn more
Enterprises facing the difficulties of placing AI agents in critical applications, the new, more pragmatic model arises, manages these people as a strategic guarantee against AI failure.
Such an instance MixusThe mission is a platform that uses “a colleague-loop” approach to make AI agents valid for critical work.
This approach is a response to the growing evidence that fully autonomous agents are a loaded gambling with high stakes.
The high price of AI not unopened
Problem AI Hallucinations AI has become a material risk as companies exploring applications. In an end event, the AI-Powered Code Editor Cursor saw his support bot To invent a fake policy Restrict subscriptions, wave of a wave of public customer cancellation.
Similarly, Fintech is famous for Klarna reverse course By replacing customer service agents with AI, he resulted in lower quality after moving. In a more exciting case, New York City’s AI-Powered Business Chatbot advised entrepreneurs to deal with illegal practiceemphasize the risks of disastrous fit of unique agents.
These events are symptoms of a greater abilities cavity. Salesforce by May 2025 Research paperToday’s leading agents emphasize a significant space between only 35% of the time in one step positions and only 35% of the time in many stairs “Current LLM opportunities and multifaceted requirements of real-world scenarios.”
Model in colleague
A new approach to make this gap bridge is focused on structured human control. “The EU agency must act in your direction and your behalf,” said Mikhus co-founder Elliot Katz Ventureat said. “But without domestic organizational control, fully autonomous agents often create more problems than they solve.”
This philosophy is upset by Mikhus’s colleague, which places the verification of man’s verification directly in automated workflows. For example, a large retailer can receive a weekly report from thousands of stores with critical operational information (eg sales, employment hours, productivity, fertility ratios, merchants). Human analysts should handle the information manually and make a decision on the basis of heuristics. With Mixus, the AI agent analyzes anomalies such as heavy lifting, analyzing complex patterns and unusual high salary requirements or productivity outs.
Decisions for high stakes such as payment permissions or policy disorders – require a human approval before continuing a “high risk” by a human user. The division of labor between AI and humans has been integrated into the process of creating an agent.
“This approach, when the experience of people actually adds the cost, 5-10% of decisions that can be significantly affected,” Katz “.” The context, judgment and accountability begins in a clear way, when the context is the most. “
The Mixus team is an intuitive process that can be done with a demo in a demo, which is shown in VentureBeat, with plain text instructions. To create a verification agent with reporters, co-founder Shai MagimoF describes the multi-step process in the natural language, and the platform can place a specific threshold and causing influential damage or legal consequences.
One of the main powerful parties of the platform, Google Drive, email and slack to enterprise users from direct contact platforms (for example, the editor connect with the direct contact platforms for the editor’s e-mail).
The integration capacity of the platform is expanding to meet special enterprise needs. Mixus supports Model context protocol (MCP), escaping the need to rediscover the steering wheel to enterprises, and allows agents to close the APIs and API (MCP). Combined with the integration for other enterprise programs such as JIRA and Salesforce, these agents allow these agents to perform complex, cross-platform tasks such as checking open engineering tickets and reducing your status slowly.
Human control as a strategic multiplier
The enterprise is currently the reality of the AI space for moving from the company’s experimentation. The consensus among many industrial leaders is a practical necessity for people in the loop to ensure the reliability of agents.
Mixus’s cooperation model changes the EU scale economy. The mixtures forecasts that until 2030, the agent will accommodate 1000x, each human controller will be more effective for each human controller to make AI agents more secure. However, the general need for human control will still grow.
“Every human controller manages more aI work in time with time, but you need more general controls than an AI placement explosion in your organization,” Katz.
For enterprise leaders, this means that human skills will be more developed. Instead of replacing the AI, experts will be presented to roles where the high stakes decisions are provided to the fleets of AI agents and are managed to review.
In this context, a powerful human control function becomes a competitive advantage that allows competitors to create competitors more aggressively and securely than AI.
“These multiplied companies will prefer the industry, and those who follow fully automation will focus on reliability, compliance and confidence,” Katz.
Source link
