A PM on the planning team started turning out structured weekly market-analysis reports after adopting an AI work tool. Before that, they used to lose hours just figuring out how to organize the same data. Once the tool entered the picture, polished output started arriving fast. The team lead said their skills had grown. The PM felt the same way.
Six months later, they joined a project with an outside partner. Someone from the partner company asked how the market-size estimate in the report had been calculated. The PM couldn't explain it — the number had gone straight from the AI tool into the report, unreviewed. Asked to produce the same report without the tool, it was harder than it had been six months earlier. They couldn't even remember where their old approach, the one from before the tool, had gone.
During the very stretch when output quality was climbing, the actual judgment had been quietly moving over to the tool. Call this shift — where the party making the call on a task moves from person to tool — judgment outsourcing. As long as the output looks good, the outsourcing stays invisible. It only surfaces when the tool is gone, or when someone has to explain the reasoning behind the result.
This piece lays out a check you can run on yourself before that gap gets exposed — what to look at at each step, and how the verdict diverges.
How to Tell When You Need This Check
Using AI tools doesn't automatically mean you need this check. It's specific signals that call for it.
The first signal to catch is stumbling when asked to explain your own output. If someone in a meeting asks "Why did you put this front and center?" or "What's the basis for this conclusion?" and you freeze — and it keeps happening — it's time to check. One caveat on how to read this: being underprepared and having outsourced judgment are not the same thing. If you're underprepared, going back through your materials lets you explain yourself. If judgment has been outsourced, going back through the materials still doesn't make the conclusion make sense, because you never actually followed the logic the AI assembled in the first place. If pulling the source material back up doesn't fix it, this isn't a prep problem.
The next signal is not knowing how to start without the tool. If you try to begin the same kind of work without AI and draw a blank on where to even start, the entry point to the task itself has been outsourced. Say a new-market analysis takes an hour with the AI tool, but without it you can't decide what to look at first — the approach isn't inside you anymore. The tool isn't just speeding up the work; it's supplying the direction. For solo operators or small teams, this signal tends to surface later, simply because fewer people are around to review the output. And a signal that surfaces late means the outsourcing has gone deeper before anyone notices.
If either signal applies to you, move to the next step.
How to Pin Down Which Judgment You Outsourced
Even once you've confirmed judgment has been outsourced, how serious it is — and what to do about it — depends on which kind of judgment it was.
Some judgments don't have much to do with your actual competence, even when handed to AI. Phrasing, sentence flow, formatting, organizing material, repetitive document formats — these are the kinds of tasks that used to go to an assistant or outside contractor anyway. The tool just does them faster. Delegating this kind of judgment doesn't erode your capability.
Other judgments you need to hold onto yourself. What to emphasize and what to cut, which direction fits a given situation, whether a stated conclusion is actually well-supported — only someone who understands the context of the work can decide these. Hand this kind of judgment to AI, and the muscle for making that same call yourself next time keeps getting weaker.
There's a way to check which kind of judgment you've been outsourcing. Pick three pieces of output from the past month. For each one, answer this:
Where did I change what the AI proposed here, and why did I change it?
If there's no record of a change, or the changes were limited to word choice or sentence length, the judgment on direction, structure, and conclusion sat with the AI. If there were changes and you can state the reason — "this market data didn't fit our company's stage, so I swapped in a different metric," or "this conclusion lands better in this order than that one" — that judgment is yours.
Check how many of the three you can give a specific reason for. If it's zero, you need the next step.
How to Check Whether Your Judgment Still Works Without AI
If the questions above raised doubts, confirm it with a direct test.
Without any AI tool, complete one core task you do often, start to finish. There's one condition: how much longer it takes doesn't count. Of course it's slower without AI. What you're judging is the character of the output.
Once it's done, set it side by side with the AI-assisted version and check two things.
First, whether the direction matches. If, using AI, you decided "this topic is the core, that one is background," did you land on the same call without it? Say that with AI you centered customer segment A, but without it you end up giving A and B equal weight — that means AI was the one deciding which one mattered more. If the direction shifted, the direction call belonged to the AI.
Second, whether the structure holds. If, with AI, you led with the claim and put the evidence after it, did you build the same structure without it? If working alone leaves you listing material with no conclusion, the structural design belonged to the AI.
If either of these shifted, you've confirmed that outsourcing is happening at that spot. This test takes thirty minutes, once a month. What matters more than the test itself is registering the result. Knowing "the direction changed once AI was out of the picture" — versus not knowing — is what determines how you use AI the next time.
How to Bring the Judgment Back to Yourself Once You've Confirmed Outsourcing
Once you've confirmed a core judgment has shifted to AI, cutting off the tool isn't a practical fix. The fix is changing the order in which the judgment happens.
Decide on direction before you open the AI tool. Before drafting a report with AI, write down first: "What does the reader need to walk away knowing? Which items are central, and which are secondary?" Make that call first, then bring in AI — now you can compare whether its output matches that direction. If AI proposes a different direction, figure out why and choose deliberately. Use AI before you've set a direction, and you end up following AI's direction without reviewing it. This one change in sequence determines where the judgment actually gets made.
Next is a habit of checking the basis for conclusions. When AI hands you a conclusion, check once: "For this conclusion to hold, what has to be true?" Once this check becomes routine, you'll be able to explain yourself later when someone asks for the reasoning. Skip it, and you freeze when the question comes. It costs a little more time. But output built by skipping this step wobbles the moment someone digs in with a hard question.
There's something that outlasts how you use AI: being able to set direction and explain the reasoning behind it, with or without the tool. That's determined not by what the tool can do, but by the attitude you bring to using it. You can't tell, just by looking at the output, the difference between someone whose judgment grew alongside their AI use and someone who just grew more dependent on the tool. It shows up only when the tool is gone — or when someone has to explain the basis for the result.
Checking yourself before that moment arrives — that's where this whole sequence points.
