What Not to Outsource to AI

AI literacy is quickly becoming a basic professional skill. LinkedIn reports that U.S. jobs requiring AI-literacy skills grew 70% over the past year, and if you care about increasing your output or simply keeping pace professionally, learning how to use these tools is becoming less optional by the month.

Somehow, we have arrived at this point without ever really figuring out social media literacy or even basic technology literacy, which is kind of insane, but I guess we keep moving.

The more important question now is not whether we should use AI, but how we can use it without gradually making ourselves less capable. The simplest rule is this: be prepared to lose any skill you consistently outsource to AI.

If AI writes all of your copy, your ability to write copy on your own will probably get worse. If it summarizes everything you read, you will get fewer repetitions at filtering useful information from noise. If it generates every idea for you, you will get less practice sitting with a blank page and coming up with something yourself.

If you don’t use it, you lose it!

That does not mean every skill is worth preserving. I do not care if a calculator makes me worse at multiplying large numbers in my head. The more useful question is which abilities you actually want to keep.

The same principle applies psychologically, and I think this is where AI gets more interesting.

Take reassurance seeking. You are unsure whether you made the right decision, so you ask ChatGPT. Then you ask whether the text you sent sounded weird, whether your boss is mad at you, or whether you should end the relationship.

Each time, you may get some immediate relief. The problem is that reassurance seeking already tends to reinforce itself. You feel uncertain, get reassurance, feel better, and become more likely to seek reassurance again the next time you feel uncertain. AI gives you access to that loop 24 hours a day.

Over time, the thing you may be outsourcing is not just decision-making. It is your ability to tolerate uncertainty, trust your judgment, regulate anxiety, and make decisions without getting someone else to co-sign them first.

So I think the rules for using AI should be built around preserving those psychological resources.

If you tend to seek reassurance, don’t ask AI what you should do. Ask it for information. Have it research the options. Make the decision yourself, then ask it to poke holes in your reasoning.

If you struggle with uncertainty, don’t use AI to make uncertainty disappear. Ask it what is known, what is unknown, and what reasonable possibilities exist. Then practice living with the fact that you still may not know for sure.

If you struggle to trust your own judgment, form an opinion before opening ChatGPT. Write down what you think first. Then use AI as opposition rather than authority: “What am I missing?” “What would someone who disagrees with me say?” “Where is my reasoning weakest?”

If you overanalyze social interactions, don’t ask AI to interpret every text. Decide what you think happened and, when it actually matters, ask the person. Otherwise you risk getting extremely good at analyzing relationships while getting worse at actually being in them.

If you avoid difficult conversations, don’t let AI have them for you. It can help you organize your thoughts or identify what you are avoiding, but you still need repetitions at saying uncomfortable things, tolerating someone else’s reaction, and repairing misunderstandings yourself.

The goal is not to use AI less. It is to be deliberate about what you refuse to hand over.

Technology is supposed to liberate us from work that does not need to consume our time. Washing machines, dishwashers, and cars do that extremely well, and I’m very happy to outsource those jobs.

AI is a different ball game because it can increasingly perform the same cognitive and psychological work that develops judgment, confidence, creativity, emotional tolerance, and interpersonal competence.

Those are not chores I’m sure we want to automate.