Have you found the power up button on your ChatGPT yet?

When I ask some of my friends what they do with ChatGPT, I hear about emails and proposals. Some ask it legal questions. Others ask about their health, or use it to understand something they have read.

Does that sound familiar? Is that what you do too?

You ask it something, it gives you an answer, and you take it from there. You copy the text, open another program, find a document, download a report. You are still the person moving everything around.

But what if you gave it access to some of those things?

Could it open your browser and click the links? Look through the folders on your computer? Find the document in Google Drive? Read the email that explains what happened with a customer?

What about QuickBooks? WhatsApp? Your social media?

Do you want it to run your life for you?

I’m being facetious, of course. But maybe not entirely. Maybe one day.

For now, there is quite a lot between asking it a question and letting it run your life. That is where I want to go with this.

And no, there is no single power up button hiding in the menu. There are tools, connections and permissions that let ChatGPT do more, and skills you can learn to make use of them.

You are giving it some control

Remember JARVIS in Iron Man? Part of the appeal was that Tony Stark could ask for something and JARVIS could work with the systems around him. That is a useful way to think about the change, with rather more setup and supervision in real life.

You are allowing the AI to take steps on your behalf. Open something. Read it. Work with it. Create something somewhere you can actually use it.

Does that feel scary? Absolutely.

We might not fully understand what is going on. But how many people know how a car works before getting into one? How many understand how an aeroplane flies? How many know where all their food comes from and how it grows?

Yet we drive to work and put our kids in the car every day. We get on a plane, travel somewhere and expect to reach the destination.

We trust the people who built those things, the people operating them and the experience we have had with them. We do not have to understand every detail before we can use them.

At the end of the day, we are human beings. Remember the Staples Easy Button? Press the button and things happen. “That was easy.” That is what we want in our lives. Simple, just like that.

Think about Amazon. How many people once thought, “Hell no, I’m not putting my credit card online”? In 2022, nearly eight in ten Canadian internet users made online purchases. You find something, order it and it turns up at your door. It is easy to see the appeal.

Sometimes we need convincing. A friend tries something and tells us about it. Then we try it ourselves. A bad news story can make us hesitate; a positive one can make us more willing to give it a go. Our confidence can change with what we hear and experience.

With AI, getting to that point takes some work. Someone has to set things up, decide what it can access and check that it does what you expect. Cars and planes have their own safety systems. With AI, we need to pay attention to the safeguards available for the task in front of us.

I explored that hesitation in Why Don’t We Trust AI? I Get It. I’m Still Using It..

Maybe someone needs to hold your hand first

And that is okay too.

I have seen it. We have done it ourselves. Someone can read about AI or hear what it can do, but then they see someone actually do it. Someone they know. Someone they have paid. A coach sitting with them, working through something they recognise.

Suddenly, it becomes easier to understand. It feels more trustworthy.

You can ask, “What did you just give it access to?” You can see which file it opened. You can watch it make something, ask for a change and see what happens next.

Someone shows us, we try it with them, and then we start doing it ourselves. You may not need that help, but it is there if you do.

That is the approach I describe in AI Is Changing Our Jobs. How Do We Learn What Comes Next?.

So what would they actually show you?

They could start with your computer and browser. In supported setups, ChatGPT can use desktop applications, open pages, click and type. Computer Use and the browser extension have separate setup and permissions. Browser access does not automatically include every application or folder.

Then there are Google Drive and Gmail connections. What it can read or change depends on the connection and permissions, including whether it may send email.

QuickBooks? Intuit documents a QuickBooks Online plugin for ChatGPT for US customers. Canadian businesses need to check availability. An exported report can also give it information to analyse.

Social media connections vary by platform and tool, so check which accounts and actions are supported. The WhatsApp Business Platform supports business messaging integrations, but that does not establish a direct ChatGPT connection to your personal WhatsApp inbox.

Those details matter. A useful demonstration includes looking at the actual permissions, not just clicking through them. AI can misread information, expose something private or change the wrong thing. You can begin with selected copies of files and a place to save new work, while keeping sending, deleting and other changes outside the assignment.

My desktop assistant article describes the experience of AI using software. The cloud and local workspace guide explains where that work happens. Here, the useful step is seeing what it can do with your kind of work.

Let’s see whether your marketing is working

What if you could bring together your customer phone conversations, the ads you are buying, the forms completed on your website and the sales made through it, all in one dashboard?

That would be amazing, wouldn’t it?

We are already working on linking these things. It is exciting to be at that frontier. I described part of this work in What I Actually Do With AI All Day. Here is an illustrative assignment:

“Use the approved call records, ad reports, website enquiries and sales exports. Build a dashboard I can filter by campaign. Show the sources behind the numbers and flag anything you cannot reliably match.”

Imagine working through that with someone beside you.

First, confirm permission to use those records, which sources it can access and where it may save the dashboard. With suitable tools, ChatGPT can read the files, calculate results and create and revise files and interactive previews.

Then you inspect what it made.

Suppose it counts every call as a new enquiry. You know one customer might call three times. So you explain:

“Use the enquiry reference to identify the same opportunity across records. Keep repeat calls separate from new enquiries. Flag anything you cannot match.”

Perhaps it treats an unanswered enquiry as a lost job.

“That one is still open. Show how long it has been waiting. Only mark it lost when there is a recorded outcome.”

Which ad gets the credit if someone clicks twice and later calls? That needs an attribution rule you understand, with uncertain matches flagged. Your knowledge of the business matters here.

You can also check enquiries you know well, compare the totals with the original reports and test the filters. A dashboard that looks convincing can still be wrong.

You may get a clearer view of which campaigns lead to sales, or discover gaps in the records. Either gives you evidence to examine.

For more on what the numbers should tell you, see Is Your SEO Working? Measure Calls, Inquiries and New Business.

And then you try it

You have seen the access being set up, the work being done and the corrections being made. You have something you can try again yourself.

As business owners, we look for useful advantages. What if my competitor figures this out before I do? FOMO. Fear of missing out. Trying one relevant task gives you a way to judge the potential advantage for yourself.

I do not know exactly where this is going. It is moving quickly. Having someone knowledgeable help you work through it can make a difference.

And what happens when your software needs to change? There is more to explore in the build-or-buy question. For now, there is plenty to explore beyond just chatting with it.

This is the approach behind our practical AI coaching. Work through something relevant together, so the next time you face it, you have a better idea of where to start.

Maybe your first step is trying a connection. Maybe it is asking someone to sit beside you and show you.

We all like the idea of the Easy Button. Sometimes it helps to have someone show us how to get there.