AI Agents: How They Save Me Time and Nearly Took Over My Computer

Table of Contents
 

• How We're Using AI to Save Time
• AI for Finding New Business
• Then This Happened...
• AI Is Powerful. That's Also the Problem.
• The Good News
• What I've Learned About AI Automation
• The Future of AI

I’ve been using AI for a while now, but recently I decided to take things a little further.

Instead of simply asking AI questions, I started giving AI bots actual jobs.

And so far, they've been pretty incredible.

We've even gone so far as to give the bots their own email address. They can receive notifications, communicate with other systems and handle certain tasks without everything having to pass through my inbox.

At this point, they're starting to feel less like software and more like employees.

I'm not sure whether I should be putting them on the payroll yet.

How We're Using AI to Save Time

One of the simplest examples is email.

I have an AI bot monitoring my inbox and helping get rid of spam, junk and the other assorted nonsense that finds its way into my email every morning.

It saves me about 20 minutes a day.

Twenty minutes may not sound like much, but that's 100 minutes a week. Over the course of a year, that's more than 80 hours.

I basically hired a robot to give me two extra workweeks a year.

Not bad.

We're also using AI automation for website maintenance. When our hosting provider sends an email notifying us that a WordPress plugin has a security vulnerability, a bot can handle the update instead of someone having to stop what they're doing, log in and take care of it manually.

That's a pretty good use of a robot.

We also use AI agents to help manage tasks from clients.

A client might send an email asking us to make a change to their website. Instead of someone manually reading the email, creating a task in our CRM, figuring out who should handle it and assigning it to the right person, an AI bot can handle much of that process.

It can read the request, create the task and assign it to the appropriate team member.

That's the kind of work AI is particularly good at: repetitive tasks that need to get done but don't necessarily require a human sitting there clicking buttons.

AI for Finding New Business

We've also started using AI for prospecting.

We can have an AI agent search the web for businesses with outdated websites, identify potential opportunities and organize the information for our team to review.

Instead of spending hours searching Google, opening hundreds of websites and keeping track of potential prospects, AI can handle much of the initial research.

And that's when I started to realize just how different AI is becoming.

This isn't just a chatbot anymore.

It's starting to look a lot more like an employee.

A very fast employee.

An employee who doesn't take lunch.

An employee who doesn't ask for a raise.

And an employee who occasionally makes you wonder if it has any idea what it's doing.

Then This Happened...

One Friday morning, I was sitting at my computer when I noticed something strange.

A browser window opened.

I didn't open it.

Then another one opened.

Then our CRM appeared.

Okay.

That's weird.

Then I noticed it was going directly to a specific task.

Task #125.

I definitely didn't open Task #125.

At this point, my first thought wasn't, "Wow, artificial intelligence is amazing."

It was more along the lines of:

"Who is messing with my computer?"

I told the AI system that the bots were affecting my computer.

It assured me that they shouldn't be. The bots were supposed to be working on their own remote computers, completely separate from my Windows computer.

I explained that I was actually seeing browser windows and mouse activity.

Then I gave it the important detail:

"My computer keeps opening our CRM."

To Task #125.

That's when things got interesting.

The AI determined that one of the bots was working on that CRM task and that its browser activity appeared to be bleeding over onto my computer.

Now, I'm generally pretty comfortable with technology.

But watching your mouse move around your computer when you aren't touching it is one of those experiences that makes you reconsider every decision you've made since purchasing the computer.

I stopped the bots.

All of them.

Immediately.

The browser and desktop activity was paused while we figured out what was going on.

Fortunately, the activity stopped.

Nobody hacked anything.

Nobody stole anything.

The robot uprising was apparently limited to one CRM task.

AI Is Powerful. That's Also the Problem.

The experience was a pretty good reminder that there is a big difference between asking AI a question and giving AI the ability to actually do things.

If ChatGPT gives you a bad answer, you can ignore it.

If an AI bot writes a terrible email, you can delete it.

If an AI agent accidentally opens the wrong CRM task, well...

You start reaching for the power button.

The more businesses use AI automation, the more important it becomes to understand exactly what these systems can access.

Can the AI read your email?

Can it send an email?

Can it access your CRM?

Can it change your website?

Can it log into your hosting account?

Can it control a browser?

Can it access your local computer?

Those are very different levels of access.

The Good News

After we figured out what happened, we resumed the bots' browser work.

But we established a very simple rule:

The bots work on their computers. I work on mine.

They can log into websites, work in the CRM and perform other browser-based tasks on their remote computers.

My computer is off limits.

Which seems like a pretty reasonable arrangement.

I'll handle the coffee and the decision-making.

The robots can handle the repetitive stuff.

What I've Learned About AI Automation

Despite the little CRM incident, I'm still a big believer in AI.

Actually, the experience probably made me more enthusiastic about it.

The amount of time these tools can save is significant. They're particularly useful for repetitive business tasks that don't necessarily require a person to sit there and click the same buttons over and over again.

For businesses of all sizes, that's especially interesting.

You don't necessarily need a huge staff or a massive technology budget to start finding ways to automate repetitive work. AI tools can help with everything from email management and website maintenance to CRM tasks, research and prospecting.

But there is an important difference between AI that gives you information and AI that can take action.

Once you give an AI agent access to your email, CRM, website, browser or other business systems, you're giving it some level of authority.

And with authority comes the possibility that it might occasionally do something you didn't expect.

That's why I'm becoming a much bigger believer in guardrails.

Know what the bot can access.

Know what it can change.

Know which accounts it can log into.

Know whether it can send something without asking you first.

And, perhaps most importantly, know which computer it's actually controlling.

Because there's a pretty big difference between:

"The AI updated a WordPress plugin for me."

and

"Why is my computer opening the CRM?"

I'm still very much in the first camp.

I'd like to stay there.

For now, anyway.

The Future of AI

I think we're still in the early stages of figuring out what AI agents will mean for businesses.

The biggest opportunity may not be replacing people. It may be giving people more time to do the things that actually require a human.

Let the AI sort through the junk email.

Let it handle repetitive website maintenance.

Let it create routine CRM tasks.

Let it research potential customers.

Then let the people on your team spend more time talking to customers, solving problems and growing the business.

That's a pretty good division of labor.

As long as the robots stay off my computer.