How many times have you gotten an email, figured it was a quick reply, and started typing? Then you stopped. Typed a little more. Copied the whole thing into ChatGPT, pasted its answer back in and hit send.
It seems harmless. It saved you five minutes and the reply was fine. But a year ago you'd have answered that email without a second thought, because you knew the person, the history behind the question and what they were really asking. Somewhere along the way, the first move stopped being yours.
The phase everyone goes through
I see the same pattern every time I take a team through AI fundamentals. The first few weeks are exciting. People see what the tools can do and start putting AI into every part of their work, from emails and reports to meeting prep, even the call on which vendor to phone back first. Once you see how strong it is, it's easy to assume it's better than you at everything.
What comes next is the part nobody plans for. People start noticing how much of their day goes to editing what AI gave them: fixing the tone so it sounds like them, correcting facts the tool had no way of knowing and adding back the context that only lives in their head. At some point it clicks that for a lot of these tasks, they'd have been faster doing it themselves from the start.
That realization isn't a step backward. It's the skill.
Not everyone gets there. Some people slide the other way, handing off one task and then another until they've stopped asking whether AI belongs in the work at all. It goes from tool to crutch to dependency, and the thinking muscle weakens so quietly that you only find out when you reach for it and it isn't there.
The real skill isn't prompting; it's discernment
Most of the conversation about AI is about how to use it: better prompts, better outputs, tighter workflows. Those help, but they won't set you apart for long, because everyone is learning the same tricks.
What separates people is knowing where AI belongs and where it doesn't. It's great for grunt work like first drafts, summaries, pulling scattered information together and routine low-stakes messages, so use it there freely. It's a poor substitute for judgment calls, high-stakes conversations or anything where the thinking itself is the value.
Here's a simple rule: if you can't explain why AI's answer is good, you likely shouldn't be using it.
Agents raise the stakes
When this was about drafting emails, the cost of outsourcing your thinking was a little rust. Agents change that.
Agents don't just write. They act. They sort the inbox, rank the leads, flag which accounts need attention and recommend what to approve. More and more, people let the agent make the call and move on.
Often the agent is right, or close to it. What it lacks is your context. It doesn't know this customer renews next month and is touchy about price. It doesn't know the "low-priority" request came from someone your CEO has known for twenty years. It doesn't know why you stopped doing it the obvious way three years ago. You know those things, and when the agent decides without you, none of them make it into the decision.
You can feed an agent some of that through account notes and the CRM, and you should. What you can't hand it is ownership of the call. When it goes wrong, the decision is still yours.
That's what outsourcing your thinking looks like now: a decision made without the person who had the context to make it well.
Better AI makes your judgment matter more
As AI gets better, the human side of the work matters more. That sounds backward until you look at what changed.
When the tools were mediocre, everyone checked the output because they had to. Now it's good enough to trust most of the time, which makes it easy to stop checking, and a confident, polished, wrong answer does a lot more damage when nobody is looking.
And when everyone has the same capable tools, the output itself stops setting anyone apart. What's left is judgment: what to do, what not to do, and what matters for this customer in this situation. You can't hand that off without paying for it.
You become replaceable.
If the thinking isn't yours, if the judgment isn't yours, if the value you add is "I ran it through AI and cleaned it up," then you're not the operator anymore. You're just the middleman. The irony is that the people automating themselves out of relevance usually believe they're being efficient.
People can tell
A lot of people think they're using AI quietly. They aren't. After a couple of years of reading AI-written email, most of us can spot it in the first few lines: the tidy structure, the polite opener that could have gone to anyone, and a reply that answers the question but misses what the person was really asking.
The people who know you notice first. A client who has worked with you for five years can tell when a note didn't come from you. They rarely say anything. They just start to wonder what else you're handing off, and how much of you they're actually getting.
That's the real cost. Using AI was never the problem; letting it replace the part of the message that was supposed to be you is.
Use AI as a thought partner
None of this means using less AI. It means putting it where it helps you decide.
The best use of AI is getting the right information in front of you at the right time, as fast as possible, so you can make a better decision sooner. That might mean pulling the account history, summarizing the last six emails or laying out the options and what each one costs, and then handing it back to the person in the chair.
For leaders, the most valuable use is as a thought partner. Before a big call, ask it to argue the other side, poke holes in your plan or tell you what a skeptical CFO would push back on. Talk the decision through with it the way you would with a well-read peer who has no stake in the outcome. It will surface things you missed. It won't know which of them matter here, and sorting that out is your job.
That's also the habit to build on your team. The question isn't whether your people are using AI. It's whether they're still the ones deciding. A team that uses AI to reach good decisions faster gets sharper every month, while a team that lets AI decide slowly becomes harder to tell apart from everyone else using the same tools.
The awareness test
Try these three questions on your own work this week, then ask your team:
1. Could I still do my job well if this tool disappeared tomorrow?
2. Am I using AI to amplify my thinking or to replace it?
3. Can I explain why the AI's output is good, or am I just trusting it?
If any answer makes you uncomfortable, you're not using AI as a tool. You're leaning on it to do your thinking.
AI is going to keep getting better at bringing you answers. That's good news, as long as you remember whose decision it is. Let it gather, draft, and make the first pass. Then sit in the chair, bring what only you know, and make the call.
That part should never be the machine's job.
