AutoGen: Under-the-Radar AI Tool That's Changing the Gamet

In This Edition:

• The Dawn of the AI Agent Team

• Introducing AutoGen

• How It Works

• A Real Life Example

• What This Means for The Rest Us

The Dawn of the AI Agent Team

Imagine if you could create your own team of AI agents to work together towards a complex task like creating next year’s sales and promotion plan in a fraction of the time it usually takes. Welcome to the dawn of autonomous AI agent teams.

Microsoft’s September Keynote

If you looked past September 2023’s Microsoft Keynote you may have noticed some groundbreaking news. Beyond all of the new announcements about Microsoft’s Copilot AI assistant, Microsoft quietly announced the arrival of AutoGen.

Introducing AutoGen

AutoGen is a free-to-the-user AI tool that allows you to create as many autonomous AI agents as you want and have them work together to do things. The power in AutoGen is twofold. First, these “agents” are customizable, conversable to you and each other, and seamlessly allow human intervention and participation. Second, AutoGen is AI platform agnostic, meaning you can integrate multiple different AI platforms and Large Language Models (LLMs) from Meta’s LLaMA to Google’s Bard and of course Open AI’s ChatGPT.

How it Works

An example of how AutoGen could work is in the not-too-distant future we could all be armed with our own autonomous team of AI agents to solve all manner of complexity in our work and lives. On a more basic level, an AutoGen team can be made up of any number of autonomous AI agents or bots with each set to its own unique task. The team will be led by a commander or “proxy” agent who can also have its own assistant agent to handle its own menial task. The proxy agent can be linked to other agents and the human user is responsible for the overall task and the human user will control if and when the proxy would provide updates on the project.

A Real-Life Example

A real-life example of how this could work is if you were a sales team leader in need of more accurate CAGR projections for your customer. Your team’s sales analyst could go into AutoGen and create a project with a lead proxy agent and several attending agents. The team’s sales analyst would pre-set the project to have the lead proxy check in with her via chat or email at noted project milestones. The proxy agent would have an assistant agent working from Meta’s LLaMa who would put together the Python code based on a deep-learning model. Given this agent’s role is so critical it would be in direct constant communication with the proxy and a third agent working from ChatGPT to check the assistant agent’s code for accuracy. There would be a fourth agent to gather and clean up the historical customer CAGR data to input into the deep learning model and a fifth agent from Microsoft’s Copilot to bring it all together into a presentable form. Finally, the proxy agent would present the findings to the team’s sales analyst who would in turn share it with the team.

The Power of AutoGen

What makes AutoGen so powerful is that when you create a team of multiple agents working together the output quality improves significantly versus a single agent or bot. Also having the autonomy to leverage multiple LLMs greatly reduces the probability of errors and the model experiencing AI hallucinations.

What This Means for the Rest of Us

While I think initially AutoGen will exclusively be a developer tool, it will be the lodestar for near-future productivity gains. As AutoGen stands today, it is a relatively easy-to-use tool for anyone who has some understanding of building a chatbot. AutoGen will follow a similar trajectory to the evolution of chatbots and how they went from being exclusively something developers could create through code to today where the average person can create a chatbot with the help of a parade of user-friendly chatbot creation APIs like Botpress and now the integration of these APIs into everyday productivity platforms like Microsoft 365.

Also, AutoGen and Autonomous AI Agents in general are light years away from replacing human teams. Like generative AI itself, they will mostly aid us in our work. In the not-too-distant future, we could all not only have our own autonomous AI agent team working for and alongside us, but we could scale up to have multiple teams doing our due diligence every day. For example, you could have the Auto AI team wake you in the morning and check your email with multiple agents reading and deciding what they can respond to and what needs your immediate attention. Another Auto AI team could be at task on checking your stock picks and using deep learning to predict future price behavior and finally, you might have an Auto AI team to help you with a pressing project at work.

Google “We Have No Moat, And Neither Does OpenAI”

Finally, this all begs the question, if it is free and so easy to use then what is the advantage versus my competition? That very question was asked and answered in Google’s infamously leaked internal Google Document that claimed open source will outcompete Google and OpenAI. The impetus behind the document was Meta’s leak of their own LLaMA LLM to the world. That event led to LLaMA becoming an open-source model that was quickly adopted by what the Google doc described as “an entire planet’s worth of free labor”.

“Data quality scales better than data size”

Regardless of whether it is open source or we all have access to the same tools, it all boils down to who owns the open-source platform. What that means for the rest of us, is if we all have access to the same tools; i.e. AutoGen, it is no different than everyone having access to spreadsheets whether it be Google Sheets or Microsoft Excel. Success in Excel is predicated on how well you are able to use it versus your peers. The basic laws of engagement remain the same. The competitive advantage will be in how quickly we adapt and choose to use AutoGen to be more competitive or as the Google document noted, “data quality scales better than data size”.

Wrap Up

That is a wrap for this edition. I appreciate your readership. If you’re interested in working with me or checking out my other work, feel free to visit my marketing accelerator’s web page at shopdragon.io and provide your contact info. If you enjoyed this edition and if you found value in it, I would appreciate you sharing it with more people. It is word-of-mouth referrals like yours that help me grow.