When to use and not use an AI agent - ChatGPT Tutorial

Before we jump in and start using one of the most powerful tools for today's workforce, AI agents, I want us to start off by thinking about when it's best to use them and when it may not be ideal to use them. I like to think of AI agents as super smart interns. They can be very intelligent, thorough, and methodical in solving problems, but if they are not managed and directed appropriately, then they may not give you the results that you're looking for. One of my favorite use cases for AI agents is taking long blocks of text or messy information and organizing it, analyzing it, and providing you with all kinds of ways to assess the data for your personal uses. A great example of this is looking at large blocks of data from customers at your organization or a huge amount of emails from people on your team and summarizing it into a way that's a lot easier for you to synthesize and respond to. You can automate this by creating a summarized report that emails to your inbox once a week, so you don't even have to think about it anymore. Another great time to use AI agents is when you have low-stakes action items to complete, but they require a little bit of nuanced judgment. Let's say you have a huge amount of customer support messages and you need to prioritize which ones are most urgent and require a quick response and deprioritize the ones that may be a little trivial and could have an automated response. Another task that AI agents could be helpful with in this context is looking at a large data set and spotting nuanced patterns that may take you hours to find if you were just sifting through that data on your own. Let's say you wanna send out a mass e-mail to different leads for your organization that could be buyers of your product. You could send a targeted mass e-mail to people based on specific things that you find when scraping data about them using an AI agent so it feels a little bit more personalized. All these use cases may sound helpful and interesting, but there are some times when AI agent workflows may not be the best tool for you to use. Specific examples of when I think it's unwise to use AI agents as a tool are when you're trying to make high-stakes decisions like financial decisions at your organization. The underlying AI tools that these AI agents rely on still have a tendency to hallucinate and I wouldn't want to add and compound that risk to my organization. Another thing you want to be careful about using AI agents for is anything that has sensitive data attached to it, like people's personal financial information or medical records, which could be compromised in the event of a security breach. That's not something that you want to give third-party AI agent tools access to, even though the more recent iterations of the agent tools that we'll be using are a lot more secure than their predecessors. And I would always say, any spreadsheets or other data from your organization that you're using to upload to an AI agent tool, I would always remove those people's names and anything that could be used to identify them, because you never know where that data might end up, especially depending on the terms of service of a lot of these different tools. And there are a lot of times when adding the extra step of an AI agent workflow would just add an unnecessary amount of complication. If all you had to do is send a one sentence e-mail to someone really quickly, it probably would be more efficient to just think of that e-mail yourself, type it up really quick and fire it off, as opposed to building out a whole workflow that you might have to oversee and correct. And ultimately this won't be the best decision for you. Once you're clear on the specific tasks that you wanna use AI agent tools to automate, I think it can be a really exciting tool. And I'll be sure to attach a handout to give you more information about ways that you can and probably shouldn't use AI agents.