Over the past couple of years, every time I’m invited to speak at an event, one of the first things I do is ask for a show of hands to see how many people have been experimenting with AI tools.

At this time last year, I got a few early adopters, then that number started to climb, and I recently looked out at a sea of hands in the air. At this point, it’s safe to assume that just about all of us have at least dabbled with one of these services. And yet, when I ask about how successful those experiments have been, the response is a collective “meh.”

Why is that? We all have access to models that have superhuman abilities, yet most of us aren’t getting a ton of value beyond mediocre blog post ideas and answers to questions that we probably could have Googled.

As with any new technology, the key is the way that we’re using it. I have to admit that I was underwhelmed at first, but then I started to learn about how these things work, and expand the way I was thinking about what they could do.

Search engines have trained us to write like cavemen. We type incredibly simple prompts in ways that we would never speak to another human. “Best hotel Toronto with pool,” or “Excel error merging data how to fix.”

AI, on the other hand, is capable of interpreting much more information, so the way that we ask it for information should more closely resemble the way we’d ask a human.

With that in mind, I set out to create a model that we could all use to get better outputs from our inputs, and the result is something that I call the TOOLS method.

The first thing to understand is that services like ChatGPT are amazing, but they don’t know anything about who we are, what we’re looking to create, or why we’re asking for this command. But when we give it that context, it instantly transforms from being a generic bot to a well-trained expert in whatever we need it to be.

Just as importantly, while it can take commands like a human, ChatGPT doesn’t have feelings. It doesn’t need to be convinced, and it won’t take offence when we give it feedback.

With that in mind, here’s how the method works:

Step 1: Train

We start by training the AI to be whatever we want it to be.

Example: “You are the world’s leading expert on digital marketing for local hotels. We are a boutique hotel located in Toronto. Our target audience is couples travelling from the North East United States.”

*Include as much detail here as possible

Step 2: Objective

Put the power of AI to work for your specific objective by telling it exactly what you want it to do.

Example: “Your objective is to launch our summer campaign to our target audience, attracting new guests and driving new bookings.”

Step 3: Order

This is where you get specific. Tell the model exactly what you’re looking for in as much detail as possible.

Example: “Create a digital marketing plan using Instagram, Email Marketing, Influencers, and any other methods that you recommend.”

Step 4: Limitations

Give it guard rails and let it know what you need it to include.

Example: “Our total budget for the campaign is $5,000. We will launch on June 1st and run through the end of July. We have one full time staff who can spend up to 20 hours per week, and access to contractors with various creative skills, as needed.”

Step 5: Style

The AI doesn’t know if you want to be inspiring, hilarious, or intellectual. You can go so far as to copy and paste your brand style guide here, or just stick to a quick description.

Example: “The style of the campaign should be fun, light, and inspire people who are considering Toronto as a destination to book a trip. We love art, creativity, and great restaurants. Our tone of voice is casual, with a friendly sense of humour, but never sarcastic or negative.”

The TOOLS method can be applied to any prompt, from generating simple images to planning out full scale campaigns, and where it really gets good is when you follow up. Remember, it doesn’t have human emotions, so don’t accept the first draft it produces and give it direct feedback.

Try it out for yourself, and let me know how it goes. You may just find yourself using ChatGPT, and other models like it, for more tasks than you would have thought it could handle.