Let's talk about AI agents - starting with their definition. AI Agents complete complex tasks without needing your input. Today, the simplest form of an agent is AI transcribers, with examples like Otter.ai, Krisp, and now Zoom's AI Companion. These bots attend video meetings as participants and summarize the key takeaways and next steps. Sam Altman, the CEO of OpenAI, recently spoke about AI agents. He talked about two AI agent camps that tech companies are racing toward.
Sam is in the first camp - where AI agents are "senior employees." You can delegate tasks to the agents. They are an "always available, always great, super capable assistant executive" and are definitively not you. Sam prefers this version because the agent can push back on you and challenge your thinking. If you ask an agent to do something questionable, it will say, yes, I can do that if you really want me to, but if I do it, here's what I think will happen. Are you sure you want me to do that?
So Sam is bullish on the "senior employee" version of AI.
Now there's the second camp. Nilay Patel recently interviewed Zoom Founder & CEO Eric Yuan. Eric is in the "digital twin" camp for AI agents.
He paints a picture of AI agents in 5 or 6 years, where they can do 90% of what today's knowledge workers can do. He sees a future where you have unlimited "digital twins." You can send them to meetings to make decisions as a practical extension of yourself. Even better, you can adjust your AI to excel at tasks you might be weak at, like negotiation. You can then send that enhanced negotiation version of yourself to a sales call.
So, what do I think about these two visions for AI?
Eric's "digital twin" version is complicated to achieve. AI may eventually understand you in your current state because they are trained on historical data and may be around you for at least the 40-hour work week. However, they are unlikely to understand how you evolve and how you handle brand-new situations. These situations and any decisions resulting from them are higher risk because they just haven't happened before. One question is, how will your digital twin seamlessly sync with you on this decision? And what happens if you disagree? It also questions our future roles at work and in society.
But this digital twin vision is crystal clear and great for motivating employees and investors to stick around. So, it's good for Zoom.
Sam's "senior manager" version of the AI could be a stepping stone to Eric's "digital twin" AI world. The senior manager AI exposes you to different angles and ideas you may have overlooked through conversation. It enables you to consider people dynamics, trust, and the long-term vision before you have the final say. The Senior Manager version is more practical and empowering and is a prime example of how human + AI will likely lead to better decisions and outcomes. You can get a feel for Sam's vision by trying the latest ChatGPT real-time conversation features.
That's my take this week on AI agents.
Share this post