The biggest mistake people make with Claude and ChatGPT is treating them like software. Treat AI like a new hire you train, and the generic answers go away.
The biggest mistake I see people make with AI is treating it like software.
I get why. Claude and ChatGPT look like software. They live in a browser tab right next to Google, and you type into a box. So people use them the way they use a search engine. They type a few words, get something back, and decide it's either magic or useless.
Here's the shift that changes everything. It's not a search box. It's a brilliant new hire.
That's not just my take. Anthropic, the company that makes Claude, says the same thing in its own guide to using Claude. It tells people to think of Claude as "a brilliant but new employee who lacks context on your norms and workflows."
What does it mean to treat AI like a coworker?
Picture a great new hire on their first day. Smart, fast, great with words. And they know nothing about your business, your customers, or how you like things done.
You wouldn't hand that person a one-line note and expect great work. You'd tell them who the customer is, what happened, and what good looks like. Then you'd look at their first try and tell them what to change.
That's exactly how you get good work out of AI. Nobody's born knowing your business. You train people. So train this.
Why do my AI answers feel generic?
Because the AI is doing what any untrained new hire would do. When it doesn't know your business, it fills the gaps with the most average answer it can come up with.
A generic hire gives you generic work. A trained one gives you work in your voice that already knows your customers. The tool is the same either way. The difference is what you've taught it.
That gap shows up in the numbers too. McKinsey's 2026 State of AI survey found nearly nine in ten companies now use AI regularly, but only about 6% get at least 5% of their profit from it. Almost everyone has the tool. Very few have trained it.
Is it about learning clever prompts?
No. You don't need special commands or wording tricks. Talk to it the way you'd talk to a capable person on your team. Clear context beats clever wording every time, and I wrote a whole piece on that called Stop Prompting, Start Briefing.
Do you work with it side by side, or hand work off?
Both, and it helps to know which one you're doing. Think about the two ways you work with any coworker.
Sometimes you pull up a chair at their desk and figure something out together, back and forth. In Claude, that's regular chat.
Other times you drop a folder in their tray with a note that says what done looks like, and you walk away. In Claude, that's called Cowork.
Start with the chair. You don't give a brand-new hire a big solo project before they know your business. Once you've trained them, handing work off makes a lot more sense.
Where should you start?
Pick two or three real things off your plate this week and run them in a chat.
- Draft an email you've been dreading, and give it the full story first, the way you'd brief a new employee
- Drop in a long document, like a proposal, a contract, or a long email thread, and ask for the key points
- Take something you already wrote and ask it to make it better
Then push back on the first draft. Tell it what to change, the same way you'd coach a person. The first answer is a starting point, not the final one.
What if it gets something wrong?
It will sometimes, same as a new hire. You're still the one reviewing the work. Check anything that matters and tell it what to fix.
That's the whole mindset. Treat it like a teammate you invest in, and it starts handing you work worth keeping!