Same job title, different paycheck

In 2019, computer vision became better than humans at reading radiology scans. The predictions were clear and confident: radiologists would disappear. Parents with children interested in medicine were told to steer them toward safer ground. Young people who might have gone into radiology were pointed elsewhere.

Today, there's a global shortage of radiologists.

Jensen Huang, CEO of Nvidia, raised this case in a recent interview with Lex Fridman. Not as a curiosity. As a lesson. The alarmists got the mechanism wrong. AI made each radiologist faster. Faster meant more scans per day, more diagnoses, more patients seen. More patients meant more hospital revenue. More revenue meant more budget for radiologists. The shortage is partly a product of AI, not despite it.

That prediction failed because it confused the tool with the purpose.

The distinction most people miss

Huang drew a clean line between what you do and why you do it. A software engineer doesn't exist to write code. A software engineer exists to solve problems, move projects forward, and find better questions to ask. Coding is the tool. Problem-solving is the purpose.

AI changes the tool. The purpose stays.

The same holds for accountants, lawyers, marketing managers, supply chain analysts, and financial advisors. The underlying reason a company employs any of these people, the judgment, the context, the ability to figure out what actually matters, doesn't change with the tools.

This is the part that gets lost in most AI-and-jobs coverage. People frame the question as: will this profession still exist? For almost every experienced professional, the answer is yes. The sharper question is: within this profession, who earns more in five years, and who earns less?

That's a different question. And the data is already answering it.

The salary split is already showing up in the numbers

According to PwC's 2025 AI Jobs Barometer, workers with high-demand AI skills earned a 56% wage premium over colleagues doing the same job without those skills. Job postings that mention at least one AI skill advertise salaries 28% higher on average than postings that don't — roughly $18,000 more per year, according to CNBC's September 2025 analysis.

These aren't premiums for AI engineers or machine learning researchers. They span non-tech roles: sales, customer service, finance, marketing, operations.

Every salary band for every professional role is starting to split. On one side: people who use AI to multiply their output, take on higher-complexity work, and deliver what used to require a team. On the other side: people doing the same job at the same pace they were doing it in 2022.

Right now, that gap is mostly invisible in published compensation surveys. Aggregated benchmarks blend both groups together, so the band appears roughly flat. Employers, though, are already pricing in the difference. When Huang was asked in the Fridman interview whether he'd hire a candidate who uses AI over one who doesn't, his answer was direct: the AI-fluent candidate, every time, for every role. Accountant, lawyer, salesperson, marketing manager. Didn't matter.

He wasn't describing a niche technical preference. He was describing professional leverage. The person who uses AI can produce more work, handle more complexity, and move faster than a peer doing things the old way. That person commands a different price in the market.

If you want to see where you sit in that split right now, upload your resume to PayScope and get a read on your current market position.

How this plays out across specific roles

The radiology example worked out at scale because the profession's purpose — diagnosing disease and helping patients — got better tools, not a replacement. The same pattern is repeating across other professions. We've covered the job-level detail in our analysis of AI exposure by occupation, but the comp implications are worth spelling out here.

Financial analysts. An analyst who can run AI-assisted scenario models, pull live market data, and produce client-ready output in a fraction of the previous time can cover more accounts. More accounts means more revenue. More revenue means that analyst's market rate gets repriced relative to peers who are still doing the same work manually.

Lawyers. Contract review, due diligence, case research — these tasks compress with AI. A lawyer who cuts 60% of the time spent on them doesn't become half a lawyer. They free up 60% of their hours for the judgment-intensive work that clients pay more for. That's a different billing profile. It shows up in comp.

Marketing managers. Content production, SEO analysis, campaign briefs, performance reporting. A marketing manager with AI can operate at a level that used to require a team of two or three. That's leverage, and leverage eventually becomes a negotiating position.

Software engineers. Huang said directly: the number of software engineers at Nvidia will grow, not decline. His reasoning tracks the radiology case. The definition of coding has expanded — writing a specification for an AI to execute is coding, and that expands the population of people who can do it, which expands the software market overall. Engineers who work well with AI tools move into higher-complexity territory: architecture decisions, evaluating AI output, designing systems. The ones who don't are holding the same position they had in 2021. Here is our analysis of software engineers' career paths.

The specifics differ by role. The structure is the same: AI compresses the lower-judgment tasks, which frees time for the higher-judgment work, which is where the comp premium lives.

Who's actually at risk

Huang was direct about where the real disruption lands. If your job is the task, not the broader purpose behind the task, you're in a more precarious position.

Data entry is the clearest example. If your entire role consists of manually moving information from one system to another, that task will be automated. There's no deeper purpose surrounding it that gets elevated by the change. The task was the job.

Most experienced professionals aren't in that position. If you've been in a field for ten or more years, your job is rarely just one task. It's judgment, context, relationships, knowing which problems are worth solving and which aren't. That doesn't get automated. It does get rewarded differently, though, depending on whether the person exercising that judgment can also use the tools that now exist to act on it faster.

We looked at this from a different angle in our piece on protecting your market value when AI disrupts your industry. The pattern holds: the professionals who adapt don't just survive the disruption, they end up on the better side of the comp split.

The split is happening inside professions, not between them.

What this means before your next review or offer

Aggregate salary data for your role is increasingly misleading. The top of the band and the bottom of the band are now occupied by people who are, in practice, doing different jobs. One group has substantially expanded what they can produce per hour. Averaging both groups and calling the result "market rate" doesn't tell you where you stand.

Before a performance review, an offer evaluation, or any negotiation, two things are worth knowing. First, where do you actually sit in your field's current pay band, not three years ago's band, the current one, for your specific role, level, and location? Second, which tier of that band does your experience and output profile point toward?

If you're using AI tools routinely in your work, you're likely producing at a level that older pay band data doesn't price accurately. That's a gap in your favor. If you're not, it's worth understanding what you're leaving on the table before someone with fewer years of experience but sharper tool fluency walks into the same conversation.

Huang's broader point was that radiology is the preview, not the exception. Nearly every profession will go through the same pattern: tools change, alarmists overstate the threat, the profession survives, but the comp distribution inside it shifts toward the people who adapted. The disruption is real. It just lands on a different variable than the one everyone was watching.

For experienced professionals, the disruption looks like this: your comp tier moves based on what you did next. Not your job title.

Before your next conversation about money, it's worth knowing exactly where you stand. Upload your resume to PayScope and get a clear read on where your profile sits in the current market.

Frequently Asked Questions

Will AI reduce the total number of jobs in most professions? The radiology case suggests no, at least not in professions where the core job involves judgment. AI made each radiologist more productive, which increased demand for the profession overall. A similar dynamic is playing out in software engineering, financial analysis, and legal work. The jobs stay. The comp distribution inside them shifts.

How much more do professionals with AI skills actually earn? According to PwC's 2025 AI Jobs Barometer, workers with high-demand AI skills earned 56% more than peers doing the same job without those skills. Job postings that mention at least one AI skill advertise salaries 28% higher on average, roughly $18,000 more per year. These premiums span non-tech roles including sales, finance, and operations.

How quickly is the salary split showing up in real compensation data? Slowly, then fast. Right now it's mostly invisible in published salary surveys because those surveys aggregate everyone in a role together. But the premium is starting to appear at the hiring stage, even if it hasn't yet shown up consistently in published benchmarks. In two to three years, it will.

If I've been in my field for over 10 years, should I be worried about AI? Probably not about your job title. Possibly about your position in the pay band. Experience gives you the judgment and context that AI doesn't replace. The question is whether you're using AI to multiply the value of that experience, or leaving that leverage on the table.

What did Jensen Huang actually say about who he would hire? In the Lex Fridman interview, Huang said he would always choose the candidate who is expert in using AI, regardless of the role. He named accountants, marketing people, supply chain managers, lawyers, and salespeople specifically. He framed it as a baseline expectation, not a differentiator.

How do I know which tier of my salary band I'm in? Start with your own profile: role, level, location, and years of experience. Then compare it against current market data, not last year's numbers. Salary data ages fast right now. PayScope runs that comparison against current benchmarks based on your actual resume, not a job title lookup.

Know what the market is paying for your exact profile before your next negotiation.

Upload your resume to PayScope and see your market position in under two minutes.