Getting Starting with AI Agents

Until agents, AI interaction was single-task oriented. 

You told it who it was (”You’re an expert tech marketing copy writer”) and gave it something to do (”take the provided background material about our company and product and develop a landing page in html for our product. We’re B2B, and the audience for the landing page is leadership.”)

Within that single task, the AI engine may do a few things (read the material, summarize it, find the parts for our audience, do its marketing thingy, and then generate html), but they were all directly related to a single request, and it has been provided with all of the information or told how to get all of the information.

An AI Agent can string a bunch of those requests together, including making decisions and finding the information it needs, like a person. 

Here’s a great example:

Prompt: “Look at my Google Calendar and find a time for a 30-minute meeting with Sarah in the next 2 weeks.”

Before agent functionality, you would have to do the following to make it work:

1. Open your Google Calendar (and Sarah’s if she shared it).
2. Check your availability (or screenshot your calendar and give it to the AI).
3. Ask the AI to draft a message for the invitation
4. You create and send that invitation (email, calendar invite, etc)

Now, the agent can do all of the following:

1. Access your Google Calendar itself.
2. Find availability.
3. Consider constraints such as travel time.
4. Ask you about preferences, such as when you typically like to meet.
5. Propose some times.
6. Actually book the meeting.

It’s now a handoff. Delegation in addition to automation. 

This is how we move toward organizational automation. 

Remember, the goal is to have AI work for you, or you’ll be working for AI.