Open any tech job listing today: 'AI specialist', 'advanced mastery of the latest tools', 'automation expert'. The problem is that this specialist does not exist, at least not in the sense that the job listing imagines. The tool that was the reference six months ago has already been replaced. The model that was state-of-the-art in January became obsolete by April. Hiring based on tool proficiency is hiring knowledge with an expiration date. And the expiration date is short.
This post shows what profile companies should be looking for, presents the criterion that almost no selection process evaluates (resistance to decision-making) and explains why the wrong candidate does not just cost a salary: it costs an invisible debt that spreads throughout the entire operation.
Hiring by tool is buying outdated knowledge
Imagine hiring a cook based on the brand of stove they know how to operate. The kitchen changes equipment the following year, and their resume loses its value overnight. What you needed from the start was someone who knew how to cook: choose the right dish for the audience, taste the seasoning, correct before serving. The stove is a detail.
With AI, it's the same, only faster. The tool replacement cycle has dropped to months. The company that writes 'advanced mastery of tool X' in the job listing is selecting based on the criterion that ages the fastest, and ignoring the criteria that remain.
This explains a phenomenon that has already appeared in several selection processes: the candidate who impresses in the interview talking about ten different tools and, six months later, is lost because the ten have changed. They memorized the manual. They did not learn to cook.
The right profile combines two skills that do not expire
The valuable professional in the AI era is not the one who masters a tool. It is the one who combines two rare skills: creativity to see solutions where the manual does not reach, and analytical thinking to break down the problem, test hypotheses, and measure what really generates results.
AI has leveled execution. Today, anyone with access to the same tools produces reports, texts, and automations with a professional appearance. What differentiates is no longer execution: it is the quality of reasoning before it. Which problem to attack. Which approach to choose. Which result to validate. This cannot be automated. It is hired.
In the automation projects that Steply implements, we see this up close: who sustains the result after delivery is not the person who knows the most tool shortcuts. It is the person who looks at the number that comes out of automation and asks 'does this make sense?' before passing it on to management.
The criterion that no selection process evaluates: resistance to decision-making
There is a mistaken idea that working with AI reduces the volume of decisions. It's the opposite: it multiplies. Each model response requires a judgment: accept, adjust, or discard? Each automation requires a choice: does this solve the right problem or just the easy problem? These are dozens of micro-decisions per day, every day.
And here appears the invisible criterion: the professional who gets tired of deciding. At first, they question everything. Over time, fatigue sets in, and they start to accept the first response the AI gives. The work continues to come out, deadlines continue to be met, and no one notices that the quality of judgment has disappeared. Until the day a big mistake explodes.
Selection processes measure tool knowledge, education, experience. Almost none measure whether the candidate can decide dozens of times a day without outsourcing judgment to the machine. And this is exactly the muscle that working with AI requires the most.
Cognitive debt: the debt that the wrong candidate brings in
Every time a professional, due to fatigue or frustration, stops judging and starts to just accept (accepts the AI response without validating, postpones the difficult decision, delegates reasoning to the tool without reviewing), they are taking a mental loan. The work comes out faster today, but the debt is recorded: unverified premises, errors that propagate, automations built on a logic that no one questioned. This is called cognitive debt, the cousin of technical debt. We have already explained how this debt arises and how Steply deals with it in projects.
And like any debt, it charges interest. The error that would cost ten minutes of review today becomes a rework of weeks when it explodes in the management report, financial closing, or proposal sent to the client. The first professional to get frustrated with the volume of decisions is the first to accumulate this debt.
A company full of cognitive debt has an easy-to-recognize symptom: AI everywhere and confidence nowhere. Everyone uses it, no one signs off on what comes out.
How to evaluate this profile in practice
The good news: it is possible to test all this in the interview, without a laboratory and without a technical test of the tool. The general rule is to ask about the decision, not the tool.
Instead of 'which AI tools do you master?', ask 'tell me about a time when AI gave you a wrong answer. How did you notice?'. Someone who has never caught AI making a mistake is not validating anything. Instead of a tool test, give a real problem from your business and observe the reasoning: where the person starts, what they ask before answering, how they decide what to measure. The tool they would use is the least important detail of the answer.
And evaluate the person's relationship with their own mistake. The professional who treats each AI output as a draft to be verified is worth more than the one who treats it as a ready answer, even if the second seems twice as fast. Their speed is a loan. The bill arrives at your cash register.
The most valuable profile in the AI era
Creative in solution, analytical in validation, tireless in decision-making. Specialist in the business problem, generalist in tools. It is hard to find, and it is exactly because of this that it is so valuable: the tool changes every quarter, the criterion does not.
The question remains for your next selection process: is your company evaluating candidates by the tool they master, or by the quality of the decisions they make when the tool fails?
