Understanding AI agents and automation

- Since this course was launched, we've had a couple of developments that can take productivity even further. Those are agents and automations. I'm not going to show you how to build these in this course. You'll find other LinkedIn Learning courses that can help you with that, but I'll give you an overview so that you've got a better understanding of these options and what they can help you with. First, let's look at AI automations. Now, automations have been around for years. It's when you connect different tools and tasks together to help you execute a process. An example would be email marketing. A process could include collecting email addresses, segmenting, scheduling pre-written emails, and tracking the open rates. The development now is that you can now send your information to AI tools as part of this process. Now, this adds a lot of power to your automations if you do it correctly. For example, you could have a spreadsheet of customer activity and use AI to write a totally relevant email to each person telling them a specific value that your new product offers them and then use AI to process the analytics and make recommendations that will help you improve opening rates. Hmm. Automation tools tend to have visual interfaces where you see how each step leads to another step. Start simple with a two-step process and start building yourself up to more powerful workflows. Next, we've got agentic AI, sometimes referred to as AI agents. The community haven't really quite settled on the definitions around these terms, you see. These are AI tools that aren't just capable of thinking. They're also capable of doing. In other words, they have agency, which is why we call them agentic. The way that you prompt these agents is by giving them a goal and all the information that they need to reach that goal effectively. Examples that are often shown are AI agents that control an internet browser to search for flights that fit your certain criteria and then taking you all the way through to the purchase page so that you just need to enter your credit card and hit buy. But they'll also be able to take on entire chunks of workplace activities like process invoices, check them against purchase orders, flag anomalies, and schedule payments. These technologies have got massive productivity implications for most companies, but I want to point out a couple of important things. The first is responsibility. Someone should be responsible for the output of an automation or an agent. Now, that means that they're responsible for the quality and the accuracy of the output. At this stage, that likely means that the output should be checked by a human. For lesser tasks, this checking might just be occasional. For vital tasks, that person becomes the filter and the editor of the output. But for some processes, you can't just leave the human involvement to the end. When I ran departments, I would ask my team to come to me with their work at various stages throughout a project, and in many cases, you should be doing the same with your automated AI workflows. This has got a name. We call it, "Having a human in the loop." When I build automations, I like to add opportunities for the user to judge preliminary output and to offer feedback before the AI continues with its process. Now, just remember that it's probably unwise to simply hand everything over to the bots at this stage because otherwise, what are you going to do? You can only play so much golf before you get bored. In fact, you can only talk about golf so much before I get bored. Oh, jeez, I'm bored already just talking about it. Let's move on to the next lesson.