Remember when ChatGPT launched and it was like, “Wow, that’s crazy. You’re telling me I can talk to a computer?” Also, remember how quickly we started to expect more from ChatGPT, like why can’t it do things for me? Why can’t it look at my calendar? Especially for our business lives, that’s super important. A white-collar employee on average touches over 10 tools a day.
Now we have these super intelligent chatbots. Of course we wanted them to accomplish things for us. But how does AI understand what to do and how to access our tools?
Remember that an AI is basically trying to respond to your prompt. In a pure chatbot context, all it has is its training data. So if you ask it “prepare me for my next meeting,” it’s going to give you generic advice based on what it understands.
Now, if you add tools to this AI (in Gumloop, you can do that by adding them directly in the UI), you can give it access to business tools and different data sources. The large language model still works exactly the same way. Even with tools, it’s still trying to predict the right next word. The difference now is that it can use these tools.
It looks at your prompt and the tools you’ve given it and tries to do the right next step. It can decide to use one of those tools. So if you say “prepare me for my next meeting,” it could decide that looking at your calendar is the right next thing to do.
What’s amazing about tool use is that large language models can figure things out on their own. They can look at your calendar to get the next meeting, then jump into your CRM to grab information about attendees, or even update information there. It takes all that information from the tools, all the meetings it’s grabbed, the success messages from the contact it created, and adds it to the prompt and decides what to do next.
Its goal is always the same: predict the right next word. But now in doing so, it can actually use the tools it has at its disposal.
There’s a lot happening in the background to allow large language models to use your tools, but you don’t have to worry about that. What’s important to remember is that with tools, you take something that you can chat with to something that can do things for you.
But here’s the thing: just because they can access your calendar, CRM, and email doesn’t mean it knows what you want to do with them. More on that in the next lesson on instructions.
Giving Your Chatbot Tools
How tools transform AI from a chatbot into an assistant that can access your calendar, CRM, email, and other business systems to take real action.
Tools are what transform AI from something you talk to into something that works for you.
Without tools, AI can only draw from what it learned during training. With tools, it can access your actual systems (your calendar, your CRM, your email, your databases) and take action based on what it finds.
The model itself doesn’t change. It’s still predicting the best next response. But now “use a tool” is one of its options.
How AI Decides Which Tools to Use
AI models are surprisingly good at figuring out which tools to use and when. When you give an AI access to multiple tools, it looks at your request, considers what’s available, and decides what to do next.
For example, if you ask “Prepare me for my next meeting,” the model might:
- Recognize it needs calendar information
- Use the calendar tool to find your next meeting
- See who’s attending
- Use the CRM tool to pull up information about those people
- Synthesize everything into a helpful briefing
The model keeps going (tool use, response, tool use, response) until it has everything it needs to give you a complete answer.
The real power isn’t just using one tool. AI can use multiple tools in sequence, taking the output from one and feeding it into the next, all to accomplish your goal.
Make the AI Work the Way You Do
Tools turn your AI from a chatbot into an assistant. But that system still has no idea how you want to work. Without guidance, every interaction with your agent may yield different results.
In the next lesson, we’ll cover how to provide instructions to your AI to help it work exactly how you want it to.