The Club of Multi-Agent Frameworks
The term multi-agent system is still young. Nevertheless, with sunrAI we have been offering a solution based on this concept for over a year and a half. Now we also welcome Microsoft to the club of multi-agent
frameworks! 🦸🦸🦸
Beginning of November 2024, Microsoft announced Magentic-One, a research project aimed at developing a "generalist agent-based system" for solving complex, multi-step tasks. It comes with AutoGenBench, an evaluation tool, now available on GitHub.
Terms like “Orchestrator,” “WebSurfer,” “FileSurfer,” “Coder,” and “ComputerTerminal” are part of the conversation. Exciting stuff! We’re thrilled to see momentum in a field we consider crucial for the future of enterprise process support.
“The future of AI is agentic”
the research team declared. We couldn’t agree more. Yet, the concept of AI agents is being approached by Microsoft, AWS, and Google as a vertical solution. That’s where the approaches differ. For tech giants, the lack of domain-specific expertise and the challenge of adapting to granular organizational needs remain significant hurdles. CIOs hoping for a plug-and-play tool like MS Excel will be disappointed.
Enter sunrAI – the Multi-Agent Platform 🦸🦸🦸 from Digitalsunray – offering solutions where Microsoft and others may still be searching for answers. Our platform emphasizes detailed process support, with a generalist framework for capturing and analyzing enterprise data, but highly specialized agents. Add to that our “Smart Data” and Agent Routines, which outperform purely generalist models.
While current research projects may not yet match the depth of real-world applications, they mark a promising direction. Here’s how sunrAI stands apart, despite similarities in the multi-agent approach:
- Agents in Charge: Magentic-One uses a single Orchestrator agent. We found scenarios at sunrAI where multiple “Agents in Charge” yield better results.
- Vendor Lock-in: Magentic-One relies on OpenAI models. sunrAI is model-agnostic, integrating with Claude, LLaMA, Mistral, and others to avoid vendor lock-in.
- Data: Data isn't a central focus for Magentic-One yet. At sunrAI, data is paramount. As #McKinsey highlights, most organizations emphasize efficiency over effectiveness. Our holistic Smart Data approach puts strategic data analysis front and center.
- Industry Focus: sunrAI provides tailored solutions for specific industries, while Magentic-One adopts a more generalized approach.
- Application Readiness: sunrAI is live and in production. Magentic-One, despite the hype, remains a research project.
You can find more details about sunrAI here.
Do you have an idea of how you want to use AI or are you still brainstorming ideas?
We at Digitalsurnay help you realize your AI project and use it strategically!
