Years of experience can make a developer faster at writing code or an analyst more efficient at building queries, but as AI takes on more of that work, speed and technical familiarity may offer less leverage in a salary negotiation.
The more valuable distinction is increasingly what professionals do with the answer: recognize a mistake, question an assumption or explain why a technically correct result would lead to a poor business decision.
“Employers now pay more for people who can tell whether an AI-generated answer is right and explain what it means for the business,” explains Vishnu Shankar, chief data officer at workforce intelligence company Draup.
Across six knowledge-work roles examined in Draup’s “The Execution Premium” report, the top 10% of earners make between two and 2.3 times as much as the bottom 10%, with the gap opening during midcareer, according to Shankar.
Those differences highlight a question for professionals approaching their next career stage: Namely, whether their experience producing expertise remains valuable when routine execution becomes easier.
Experience Needs to Produce Judgment
Shankar points to data analysis as an example: Writing SQL quickly once helped distinguish an experienced analyst from a newcomer. AI can now produce queries rapidly, making the ability to evaluate their relevance and accuracy more important.
“Somewhere between years five and 10, some professionals start making calls that depend on context that no dataset holds,” he says.
An AI system might identify unusually high attrition on an engineering team, for example. Determining whether the cause is compensation, management or stalled promotions requires knowledge of the organization and its people.
Dr. Gina Smith, senior research director for IT skills for digital business at IDC, says she sees a similar distinction between procedural expertise and knowledge developed through practice.
“The top earners will be people who pair deep domain knowledge with the ability to direct and check AI,” she says.
In IDC’s March 2026 Future Enterprise Resiliency and Spending (FERS) survey, 52% of leaders identified analytical judgment as the most essential human skill for daily work with AI agents, according to Smith.
That does not make technical knowledge disposable. Professionals need enough depth to recognize when an output is misleading or incomplete. The difference is whether they can apply that knowledge to unfamiliar situations.
“When I evaluate talent, I look for curatorial taste,” Shankar says. “Give someone two outputs from an AI tool, and the people whose earnings keep climbing can tell you which one is more accurate or useful and why.”
Build Evidence Before Asking for More Money
Demonstrating judgment takes more than adding “critical thinking” to a resume. Professionals need examples showing how their intervention changed an outcome.
“In a pay conversation, the stronger evidence is a record of decisions where your judgment changed an outcome, such as a security flaw you caught in AI-generated code before release or a design call that kept a system stable as traffic grew,” Shankar says.
Smith says she recommends keeping a record of rejected AI outputs, the reasons for rejecting them and the consequences that could have followed. Other useful measures include avoided rework, faster incident resolution and costs or revenue associated with a decision.
“Keep a log of AI outputs you rejected, why you rejected them and what they would have cost,” she suggests. “That turns judgment into a number a manager can defend at budget time.”
Any estimated savings should have a defensible basis. An identified error is evidence; a speculative claim that it prevented a multimillion-dollar loss is harder to support.
Professionals should also connect their examples to their employer’s priorities. A team struggling with release reliability may value fewer defects more than higher code output, while a finance function may care more about accurate forecasts than faster reporting.
Learn the Business Behind the Request
Communication becomes valuable when it helps someone act on technical work.
Draup found 21 data analyst profiles listing SQL for every one listing data storytelling, Shankar notes. While that’s a comparison of listed skills, rather than proof that everyone omitting storytelling lacks it, it suggests an opportunity to demonstrate capabilities beyond familiar tools.
“From there, start every analysis with the decision behind it,” he says. “Before running a query, find out what the person asking plans to do with the answer and what would change their mind.”
That approach helps analysts test whether they are answering the right question before producing a polished result.
Shankar says he recommends spending time with the business teams that own the work and presenting findings directly to sales or finance leaders. Their questions and objections can reveal what an analysis failed to explain.
Early-career professionals also need opportunities to observe experienced colleagues making decisions. As AI absorbs repetitive assignments, some traditional ways of developing business knowledge may become less available.
“One way is to shadow the mid-level colleagues who manage AI agents—ideally on more than one team,” Shankar says.
Watch for a Plateau Before It Arrives
Smith identifies several warning signs: work dominated by repeatable tasks, a core skill that AI now supplies to junior colleagues, and performance measures focused entirely on output volume.
Other signals include going a year without learning a new tool or domain and being excluded from discussions about the organization’s AI rollout.
“Build analytical judgment and learning agility first,” Smith says, noting they ranked as the top two essential skills for working with AI agents in IDC’s FERS survey.
“They also ranked as the two most underdeveloped,” she adds.
For professionals, the practical response is to seek assignments involving ambiguity, business context and responsibility for outcomes—and document what they learn.
Employers must also provide opportunities to exercise judgment and recognize it in advancement decisions. Asking workers to develop new capabilities while continuing to reward only production volume sends a conflicting message.
“Employers also need to factor judgment and communication into promotions and pay,” Shankar says. “If raises only reward technical output, employees will keep investing in the work AI is taking over.”
