Forget Siri. Your Next Assistant Will Be an AI Agent

Welcome to Leading Disruption, a weekly letter about disruptive leadership in a transforming world. Every week we’ll discover how the best leaders set strategy, build culture, and manage uncertainty all in service of driving disruptive, transformative growth. For more insights like these, join my private email list.

When I was at Forrester, I worked with a fantastic researcher.?

I remember walking up to her desk time after time and saying, “I think we should do this.” And her response was usually something like: “I was thinking the same thing! I already started on it.”

She could anticipate the next steps, produce them, and then deliver an excellent first draft.

Just imagine having someone like her by your side, always anticipating your needs and looking around corners for you.

That possibility might not be too far off! On Tuesday’s livestream, I shared why I think everyone will have a generative AI personal assistant within the next year.?

But:

What are autonomous agents — and why do you need one?

Autonomous agents are AI agents that use the large language models behind ChatGPT and other generative AI tools to carry out complex tasks on your behalf.?

They break down complicated tasks into subtasks and execute them without you sitting there hitting buttons. It’s like having your own research associate!

These agents understand the context and environment of your requests, make decisions, take actions to achieve your goal, and (importantly!) take additional action. They effectively rationalize: “That wasn’t the best answer. It changes these other things. Now I have to think about this other option and consider these things.”

If you’ve experimented with ChatGPT, you’ve noticed it’s not connected to the internet. It can’t go and look for flights, for example. It can’t feed you real-time information. It relies on other tools to conduct those searches. And it can’t access your personal information, like your frequent flier miles, to help you decide which airline to book your flight on.?

But an AI agent can use various APIs to access information, break down tasks, and run them autonomously.?

This is different from Siri!

Autonomous agents are already around. You probably have one on your phone, whether it’s Siri or Alexa.?

But these agents aren’t that intelligent. They’re task-oriented and can only carry out one function well. Siri is amazing at setting a timer, but that’s a simple command.?

And autonomous agents are more advanced than the AI used on social media. The feeds you see on Instagram, for example, are pre-programmed against a particular algorithm that makes judgments based on the information already present.?

AI agents are much more flexible than that! And they’re working very differently.?

They’re auto prompting — they’re taking complex tasks, breaking them down into subtasks, and prioritizing the information into steps. AI agents have access to tools, like GPT plugins, so you can run a search. And they improve on themselves, looking for additional information you might need.

How to take advantage of AI agents…today?

Given the pace of development (BabyAGI is barely a month old!), I believe we’re all going to have agents working on our behalf within a year.?

And that means we need to learn how to use them. Here are four tips to get you started:

1. Be specific. In the beginning, you’ll need to be incredibly explicit about the task. You’ll need to break the tasks down into as much detail as possible. But, as the agent learns from you, you can give it less and less detail, and it will work even better — without you sitting in front of the keyboard running commands.
2. Ask good questions. Using an AI agent requires a different skill set. You need to know how to query it, especially now, when these agents are new and need a lot of handholding. But we all learned how to run effective searches, so we can learn how to use this technology effectively, too.
3. Play around! Practice using these agents so you have an early mover advantage. You’ll learn how complex the tasks can be and how much you can push the AI. The more you experiment, the better the agents will get at discerning the underlying meaning behind the tasks. I encourage you to try out Godmode.space and Cognosys.ai, the two most user-friendly tools available right now.???
4. Brainstorm uses. Our imagination is constrained because we’re used to search queries. We run a search and piece everything together. But if we take away that constraint and have an agent working on our behalf, we can remove that limitation. So think about the use cases you want. If an AI agent could do anything, what would you want it to do? This is an opportunity to get creative!

With any AI tool, it’s our job as humans and leaders to control the technology. We need to understand what to ask and how to put restraints on it. And we need to consider what responsible and safe use looks like because these tools are only as good as the content we put in.

Those are some of the questions we’ll dive into next week. Join me on Tuesday, May 2 at 9 am PT to learn what organizations must do to stay safe when deploying AI. I’ll see you then!

Your Turn

What would you want an AI agent to do for you? What complex task would you ask it to tackle? I’d love to hear what you’d like to see! Let me know in the comments.