AI should be the second voice in your work, not the first. Here is how to keep it that way.
There is a quiet risk in getting good with AI. The more capable the tool, the easier it is to let it lead, to prompt first and think second, until your own judgment fades into the background. The professionals who thrive keep AI in the passenger seat. Here are five habits that keep you in the driver’s seat.
The phrase worth keeping in mind is co-intelligence: AI working alongside human judgment, not in place of it. That only happens on purpose. Left to default, the convenience of a fast, fluent tool nudges you toward letting it decide, and the skills you are supposed to be building quietly stall. One of those habits is simply checking the work before it goes out. Small moves, but together they are the difference between using AI and being used by it.
1. Think before you prompt
Form your own view first. Before you ask AI for an answer, spend two minutes deciding what you actually think the answer is. Then use AI to challenge it, extend it, or find what you missed. This small reversal changes everything. You stay the source of the idea, and the AI becomes a tool that sharpens your thinking rather than a vending machine that replaces it. It also makes you much better at catching the AI’s mistakes, because you walked in with a view of your own to compare against.
2. Own the output, always
When your name is on the work, the judgment is yours and so is the accountability. The tool is never the excuse. The lawyer who submitted AI-invented case citations could not hand the blame to the software, and neither could the firm whose report contained fabricated sources. Adopt the rule early: if you would not put your name behind it, it is not ready, no matter how it was produced.
3. Verify before you trust, especially when it looks good
The most counterintuitive finding in recent AI research is that polished output gets less scrutiny, not more. When Anthropic studied how people actually use AI, it found that the better and more finished a result looked, the less people questioned it. Flip that instinct. Treat a clean, confident answer as a reason to check harder, not a reason to relax. Human first means your judgment is the final check, never the AI’s confidence.
4. Keep the human moments human
Use AI for preparation and drafts, not for the moments that depend on presence and empathy. Prepping talking points for a hard conversation is fine. Outsourcing the conversation itself is not. A thank-you, an apology, feedback to a colleague, a check-in with someone who is struggling: these are where being a person is the entire value. Protect them, and people will trust you with the things that matter. The same logic applies to your own standing at work. The relationships you build and the trust a manager places in you do not transfer to a tool, so show up as yourself in the moments that build them.
5. Keep learning the skill underneath
This is the habit that matters most over a career. If you let AI do a task you never actually learned, you lose the ability to judge whether it did the task well. Research on the jagged frontier of AI capability found that people who simply adopted AI output without engaging tended to perform worse, because they could not tell when the tool had wandered off course. Your developing expertise is what lets you supervise AI instead of being led by it. And as the labor market reshapes around these tools, that human judgment is becoming more valuable, not less. Do not let the tool quietly erode the very skill that makes you employable.
What this looks like in practice
Picture a normal Tuesday. You open a task and jot your own rough take before touching any AI tool, then ask the AI to poke holes in it. You use it to draft a client email, but you read every line, rewrite the two sentences that sound generic, and confirm the one factual claim before you send. When a teammate seems frustrated, you close the laptop and talk to them rather than generating the perfect message. By the end of the day you have moved faster than you would have alone, and every decision still ran through you. That is the entire model. Not less AI. More ownership.
Stay in the loop
Human first, AI second is not about using AI less for its own sake. It is about staying the professional in the loop: the one who thinks first, owns the result, checks the polished answer, knows when to set the tool down entirely, and keeps getting better at the underlying craft. Used that way, AI amplifies your judgment instead of quietly replacing it.
