Consider an illustrative case. A senior marketing manager, eleven years into consumer brands, has two browser tabs open on a Sunday evening. One is a nine-month certificate in machine learning, sold as a route into “AI jobs”. The other is a posting for her own job title at a competitor. Two years ago that posting asked for channel strategy, budget ownership and a record of growing a brand. It still does. It now also asks for experience redesigning campaign workflows with AI tools, and for judgment about where those tools should not be trusted.

Her question, whether to retrain into AI, is one a great many experienced professionals are asking. The evidence suggests it is usually the wrong question. Dedicated AI roles are growing quickly, they pay well, and they are real. They are also a small share of the market, and they are mostly filled by experienced people. The larger and nearer change is happening inside existing jobs, including hers. For most people, the move that pays is to add AI depth to expertise they already have, not to chase a new title.

The new roles are real

Start with the case for the new roles, because it is a strong one. US job postings that the labor-market analytics firm Lightcast classifies as AI-related numbered 9,241 in January 2022. By August 2026 there were 80,597, an increase of 772% [1]. PwC, analyzing more than one billion job advertisements across 27 countries and territories, finds that jobs requiring specific AI skills grew 69% over the period it studied, almost eight times faster than the jobs market as a whole, at 9% [2].

They pay. PwC puts the average wage premium for AI skills at 62% in its 2026 analysis, up from 57% a year earlier, and as high as 118% in consumer markets [2]. Lightcast’s estimate, drawn from more than 1.3 billion postings, is lower: 28%, or nearly $18,000 a year [3]. The gap between the two figures is itself instructive. Both compare advertised pay in postings that ask for AI skills with postings that do not. Neither follows a person before and after they learn something. A premium in job advertisements tells you what employers will pay for a combination of skills; it does not tell you what a certificate will add to your salary.

Some specific roles have grown at striking rates. LinkUp, which collects listings directly from employer websites, counts listings for forward-deployed engineers, who work alongside customers to put AI systems into production, rising from about 300 to more than 5,000 [4]. The US Bureau of Labor Statistics projects employment of data scientists to grow 35% between 2025 and 2035 [5].

And they are few

Fast growth from a small base flatters. LinkUp’s 5,000 forward-deployed engineer listings sit in a US market that the same firm measures at 7 to 8 million active listings [4]. Data scientists held about 275,600 US jobs in 2025, and the BLS expects about 24,800 openings a year over the coming decade [5]. Those are good numbers for a profession and small ones for a labor market.

Even the broadest measures stay in single digits. AI skills appeared in about 2.5% of US job postings in 2025, according to Lightcast’s analysis for the Stanford AI Index, up 55% in a year [6]. Indeed’s tracker, which counts any posting that mentions AI, reached a high of 4.2% in December 2025 [7]. And most of the postings inside those percentages are not for AI specialists. They are ordinary jobs that now mention AI, which is the point this article will come back to. By any of these counts, roles built around AI itself are a small share of the openings available.

And they favor experience

The new roles also sit in the part of the market that favors experience. Across the US, only about 14% of postings on Indeed are for senior positions, defined as five or more years of experience. But senior postings were up 14.7% year over year in May 2026, while entry-level postings fell 7.5% [8]. In software development, where much of the dedicated AI work is classified, senior positions made up 69.3% of postings in the first quarter of 2026 [8], and 37% of the recent growth in software postings came from jobs with AI in the title [9]. Those figures do not isolate the seniority of the AI roles themselves, but the field they sit in is heavily weighted toward experienced people. Where the data are specific, they point the same way: LinkUp says of forward-deployed engineering that it is “not a landing spot for entry-level workers” [4].

Even jobs labeled entry level are being written for people with experience. PwC’s analysis of 2.4 million US entry-level postings finds that AI-exposed entry-level roles are seven times more likely than others to require skills traditionally associated with senior staff [2].

For an experienced professional, that cuts two ways. The new roles are not closed to you because of your tenure; employers are paying for judgment. And if the premium is on judgment, it is likely to attach to judgment about the work AI is being applied to. That suggests a product manager with a decade in payments who has shipped an AI feature is a more natural candidate for an AI product role than a career changer with a credential and nothing shipped.

The larger shift is inside existing jobs

The part of the data that gets less attention is the most useful. Indeed counts the US job titles that employers have written “AI” into, with enough postings to rule out one-offs. There were 264 such titles in 2022, 2.6% of the titles it tracks. By the first quarter of 2026 there were 822, about one title in twelve, and 63% of them were outside technology occupations [10]. Its examples make the pattern plain. A “Physical Therapist (AI Documentation)” is a physical therapist, with a new tool built into the job.

The same pattern runs through job descriptions. In December 2025, 45% of data and analytics postings on Indeed mentioned AI. In marketing the share rose from 8.4% to 14.9% over the year. In human resources it doubled, from 4.4% to 8.8%. In accounting it was still only 6% [7]. Lightcast finds that just over half of postings requiring AI skills, 51%, were for occupations outside IT and computer science [3].

The requirements are also moving quickly. PwC found in 2025 that the skills employers seek were changing 66% faster in the occupations most exposed to AI, up from 25% a year earlier [11]. Employers say they intend to act on it. In Microsoft and LinkedIn’s survey of 31,000 knowledge workers in 31 markets, 66% of leaders said they would not hire someone without AI skills, and 71% said they would rather hire a less experienced candidate with AI skills than a more experienced one without them [12]. That is a statement of intent, not a record of hiring, and a question this broad does not say how high a bar “AI skills” sets; it is safer to read it as a baseline than as a differentiator. Even so, it is a direct warning to anyone relying on experience alone. The same survey found that only 39% of people using AI at work had received training from their employer [12]. For most people, the depth will have to be built on their own initiative, inside the job they already have.

Why depth in your own field pays more often

The strongest argument for deepening your own field, not leaving it, comes from what AI does to the value of experience. A Dallas Fed analysis of wage growth across occupations since 2022 finds that where experience commands a large pay premium, greater exposure to AI is associated with faster wage growth, and where experience earns little, it is associated with slower growth. The author’s reading is that AI can automate codified, book-learned knowledge but not the tacit knowledge that comes from doing the work [13]. The argument that experience is being repriced, not erased, is set out at length in the article on career capital after automation.

PwC’s 2026 analysis points the same way. It separates “professionalised” roles, in which AI acts as a force multiplier for experts, from “democratised” roles, in which AI makes the work easier for non-experts to do. The professionalised roles are seeing twice the growth in available jobs and 42% faster salary growth [2]. The market is paying most for expertise that AI amplifies.

Retraining from scratch has a weaker record. A New York Fed study of federally funded training found that AI-exposed workers who retrained toward AI-intensive occupations earned returns 29% lower than similar workers who took more general training [14]. The study covered lower-income displaced workers, mostly before generative AI, so it is not a verdict on a senior professional’s certificate. But it is a caution against treating an AI credential as a destination in itself, a point taken further in the article on roles that are contracting.

Reading the market for your own role

The national data tell you which way the market is moving. Your decision turns on what is happening to your own role, and the most reliable reading comes from the people hiring for it. Treat current job postings as a market-sensing exercise: a structured comparison of what employers ask for at your level and the level above, which takes an afternoon.

  • Pull twenty current postings for your title, at your level and one level up, in your region or open to remote candidates. Count how many mention AI at all, and note where: the title, the responsibilities or the requirements.
  • Sort what is being asked for. A useful working division has four kinds of AI work: using AI tools within the existing job; redesigning a process or workflow around them; checking, evaluating or governing what they produce; and building or deploying AI systems. The first is fast becoming a baseline. The middle two are where people with domain expertise hold the advantage. The last is the specialist’s territory.
  • Compare with the level above. If the senior version of your job asks for workflow redesign or governance and yours asks only for tool use, you have found the gap that matters for your next step.
  • Look sideways at titles that now carry AI. Indeed groups the fast-growing AI-titled jobs outside tech into three clusters: enablement and consulting, training and content creation, and instruction [10]. Each is an existing kind of expertise with AI attached, and each may sit one step from your current role.
  • Repeat the exercise every six months. Requirements are moving fast enough that last year’s reading may already be out of date.

Building depth that others can see

Depth counts only when someone else can see it. If employers want experience of applying AI to real work, the asset to build is a record of having done so: the problem, the approach, what changed and how you know.

  • Choose one consequential, recurring piece of your work and redesign it with AI from end to end, not just a single step within it.
  • Measure before and after. Time, error rate, cycle time, cost or quality, whichever your function already tracks. A number that your manager already watches is worth more than a new one you invent.
  • Record where the tool failed and how you caught it. Evaluation and judgment are the scarce part of the work, and the part a hiring panel or promotion committee is trying to assess.
  • Teach it to someone else. A redesign that others adopt is evidence of influence as well as skill, and it marks you as the person in your function others ask.

The discipline of capturing that evidence as it happens is covered in the article on documenting impact. Persona Map is a career intelligence platform built to hold it: a private workspace where each project, its result and the people who saw it are kept as they happen, so the record is ready when a review, a promotion case or an interview asks for it.

When a dedicated AI role does make sense

None of this makes the dedicated roles a mistake. For some people they are the obvious next step, and the evidence above suggests three conditions under which the move tends to make sense.

  • You are already doing the work. If you have spent the past year putting AI into use in your function and others now come to you for it, an enablement, solutions or AI product role formalizes what you do. The jump is short, and your record already makes the case.
  • Your domain is the job. Forward-deployed and applied AI roles exist to connect a model to a business problem. If your expertise is in the problem, whether that is insurance claims, supply chains or clinical documentation, you hold the scarcer side of that connection.
  • You can make the move from the inside. A new role next to your own, in an organization that already knows your work, carries less risk than an external move into an unfamiliar title where you must prove both the domain and the technology at once.

If none of the three applies, the better bet is usually the one already in front of you: become the person in your function who uses AI best, and who can show it.

Back to the two tabs

The certificate in the illustrative case is not worthless. But the posting for her own job is the better guide. It tells her that the market will pay for a marketing leader who can redesign the work with AI and judge the result, and that before long it will expect it. The new AI roles are real, and for some people they are the right move. For most experienced professionals, the larger and nearer opportunity is to become the expert in their own field who works with AI better than anyone else, and to have the evidence to prove it.

Sources

  1. CFO.com. “Demand for AI Expertise Far Outpaces Wage Growth.” Report on Lightcast job-posting data. September 24, 2026. Link
  2. PwC. “AI Reshapes Global Labour Market into Two Distinct Paths, Rewarding Human Skills.” 2026 Global AI Jobs Barometer, press release. June 15, 2026. Link
  3. Lightcast. “New Lightcast Report: AI Skills Command 28% Salary Premium as Demand Shifts Beyond Tech Industry.” Beyond the Buzz, July 2025. Link
  4. LinkUp. “Inside the AI Revolution: LinkUp Job Data Shows Steady Hiring Overall, with a Different Mix of Jobs.” July 2026. Link
  5. US Bureau of Labor Statistics. Occupational Outlook Handbook: Data Scientists. Link
  6. Lightcast. AI skills in job postings, analysis for the Stanford AI Index 2026. Link
  7. Indeed Hiring Lab. Cory Stahle. “January 2026 US Labor Market Update: Jobs Mentioning AI Are Growing Amid Broader Hiring Weakness.” January 22, 2026. Link
  8. Indeed Hiring Lab. Felix Aidala & Sneha Puri. “The Labor Market Is Tilting Toward Seniority.” July 23, 2026. Link
  9. Indeed Hiring Lab. Guillermo Gallacher. “AI and Job Postings: From Destruction to Creation?” July 8, 2026. Link
  10. Indeed Hiring Lab. Pawel Adrjan. “AI Is No Longer Just a Tech Occupation Story.” July 8, 2026. Link
  11. PwC. 2025 Global AI Jobs Barometer, press release. June 3, 2025. Link
  12. Microsoft & LinkedIn. “AI at Work Is Here. Now Comes the Hard Part.” 2024 Work Trend Index. May 8, 2024. Link
  13. Federal Reserve Bank of Dallas. J. Scott Davis. “AI Is Simultaneously Aiding and Replacing Workers, Wage Data Suggest.” February 24, 2026. Link
  14. Federal Reserve Bank of New York. Staff Report 1165, on returns to retraining for AI-exposed workers. 2025. Link