Artificial Idea | AI careers · practical prompts · no hype Monday, November 24, 2025 · Issue #33 · Jobs
The salary gap
The compensation data for 2025 is in. The gap between AI-fluent and non-AI-fluent professionals in equivalent roles is larger than last year, larger than the year before, and showing no sign of narrowing.
In October 2025 the Burning Glass Institute released the most comprehensive compensation analysis of AI skill premiums conducted to date, covering 4.2 million job postings and 890,000 actual compensation outcomes across seventeen countries and thirty professional functions. The report's headline finding received moderate coverage. The detail underneath the headline received almost none.
The headline: professionals in roles requiring AI skills earn an average of 23% more than professionals in equivalent roles not requiring them.
The detail: that 23% average conceals a distribution that is far more instructive than the average, a set of sector and function-specific findings that change the picture significantly depending on where you sit, and a trajectory analysis that is more consequential than the snapshot figure for anyone making career decisions over a multi-year horizon.
This issue covers all three.
The distribution behind the average
The 23% average premium is real and well-supported. It is also the least useful number in the report for a professional trying to understand what it means for their specific situation, because the distribution around that average is wide and the factors driving the width are identifiable and actionable.
At the lower end of the premium distribution, professionals in administrative and operational functions where AI tool usage is becoming standard and the relevant tools are accessible to anyone with a browser are seeing premiums of 8 to 12%. The premium is real but modest, reflecting the fact that the AI skill in these contexts is becoming less scarce as adoption widens. A customer service professional who uses AI-assisted response tools is earning more than one who does not, but the margin is compressing as the tools become ubiquitous and the capability becomes baseline rather than differentiated.
At the upper end, professionals in functions where AI fluency requires genuine domain expertise to apply effectively, where the outputs of AI tools require expert judgment to evaluate, and where the combination of domain knowledge and AI capability is still genuinely scarce are seeing premiums of 35 to 52%. Financial analysts using AI for complex scenario modelling. Senior marketers using AI for strategic audience analysis. Legal professionals using AI-assisted research with the expertise to evaluate output quality. Clinical professionals using AI diagnostic tools with the medical judgment to assess recommendations. In these contexts the premium is large, growing, and structurally connected to the scarcity of the combination rather than to the AI skill in isolation.
The practical implication of this distribution is the one this newsletter has been building toward since Issue #5: the combination of deep domain expertise and genuine AI fluency is the position that commands the highest premium, and it is the position that is hardest to replicate quickly because it requires both the domain expertise and the AI fluency to be developed to a level where they genuinely interact and amplify each other.
A shallow AI skill applied to shallow domain knowledge produces a small premium. Deep AI fluency applied to deep domain knowledge produces a large one. The investment logic follows directly.
The sector breakdown
The Burning Glass data includes sector-specific premium figures that are worth examining for any professional whose career is concentrated in a particular industry.
In financial services, the premium for AI-fluent professionals is 31% on average, with the highest premiums concentrated in risk analysis, quantitative research, and financial planning functions where AI augments rather than replaces the analytical judgment that defines the role. The premium has grown from 19% in 2023, a 12 percentage point increase in two years that reflects both the deepening adoption of AI tools in the sector and the still-limited supply of professionals who can use them at the level the sector's analytical requirements demand.
In professional services, specifically consulting and legal, the premium is 28% on average with significant variation by seniority. Junior professionals in these sectors see premiums of 18 to 22%, reflecting the compression of entry-level work described in Issue #11. Senior professionals see premiums of 34 to 41%, reflecting the scarcity of professionals who have both the domain authority to make high-stakes recommendations and the AI fluency to make those recommendations faster, with better information, and with higher analytical rigour than was previously possible.
In technology, the premium structure is more complex because the baseline expectation of AI engagement is higher than in other sectors. The premium for AI skills among software engineers, for example, is only 14%, because AI tool usage is expected rather than differentiated at this point in most technology organisations. The premium in technology is concentrated not in AI usage per se but in the specific capability to architect and evaluate AI-augmented systems, a capability that commands a 47% premium and that sits at the intersection of software engineering expertise and AI systems understanding.
In healthcare administration, a sector this newsletter has not addressed in detail and that deserves specific treatment, the premium is 26% and growing at the fastest rate of any sector in the Burning Glass data, at 9 percentage points of growth in the past eighteen months. The drivers are the regulatory complexity of AI in clinical contexts, the scarcity of administrators who understand both the operational requirements of healthcare settings and the specific risk and compliance requirements of AI tool deployment in those settings, and the accelerating pressure on healthcare organisations to use AI to manage cost and capacity without creating clinical risk. The combination of skills required to do this well is exceptionally scarce and the premium reflects it.
The trajectory is more important than the snapshot
The 23% average premium for 2025 is the fourth consecutive annual reading in the Burning Glass longitudinal dataset. The 2022 reading was 11%. The 2023 reading was 17%. The 2024 reading was 20%. The 2025 reading is 23%.
That trajectory is not linear. It is accelerating. The year-on-year increase has grown from 6 points between 2022 and 2023, to 3 points between 2023 and 2024, and back to 3 points between 2024 and 2025. The acceleration is not dramatic but the compounding is. A premium that grows at an average of 4 percentage points per year from a base of 23% reaches 39% in four years. At that point the compensation difference between AI-fluent and non-AI-fluent professionals in equivalent roles is not a career consideration. It is a standard of living consideration.
The trajectory also has geographic variation that matters for this newsletter's readership. In India, the Burning Glass data shows a 2025 premium of 27%, higher than the global average, growing from 14% in 2022. The rate of premium growth in India is faster than the global average, reflecting both the rapid adoption of AI tools in the Indian professional services and technology sectors and the still-limited supply of professionals who have developed the combination of domain expertise and AI fluency that the upper end of the premium distribution rewards.
The professionals positioned at the upper end of that distribution in India in 2025 are those who recognised the trajectory early and invested accordingly. The professionals who can position themselves there in the next two to three years are those who recognise it now and act with the same logic.
What the premium buyers are actually paying for
The most useful section of the Burning Glass report for career decision-making is not the premium figures. It is the job description analysis that identifies what specific capabilities the employers paying the highest premiums are specifying when they advertise for them.
The top five capabilities appearing in postings at the 35% premium level and above, in order of frequency, are the following.
First, the ability to evaluate AI-generated outputs for accuracy and appropriateness in domain-specific contexts. This appeared in 67% of high-premium postings. It is an analytical capability, not a technical one.
Second, the ability to design workflows that integrate AI tools into existing professional processes without creating new risks or dependencies that reduce rather than increase organisational resilience. This appeared in 54% of high-premium postings. It is a systems thinking capability.
Third, the ability to communicate AI-assisted analysis and recommendations to senior stakeholders who are not AI-fluent, in ways that convey both the insight and the appropriate uncertainty. This appeared in 49% of high-premium postings. It is a communication capability.
Fourth, the ability to identify the limits of AI tool reliability in specific professional contexts and to maintain appropriate human oversight at the points where tool limitations create risk. This appeared in 43% of high-premium postings. It is a judgment capability.
Fifth, domain expertise at a level sufficient to direct AI tools toward problems the professional understands deeply enough to evaluate the quality of the output. This appeared in 39% of high-premium postings as an explicit requirement, and in effectively all of them as an implicit one.
None of these five capabilities is a technical skill. All of them are the capabilities this newsletter has been arguing for since Issue #1. The Burning Glass compensation data provides the empirical validation for that argument in the specific terms of what the market is paying for it.
The action
Return to the five capabilities listed above. Evaluate your current demonstrable level in each one, honestly and specifically, not in terms of how you would describe yourself in an interview but in terms of the specific evidence you could point to.
The gap between your current position and a credible claim to all five capabilities is your development roadmap. It is more specific than any generic AI upskilling recommendation and more directly connected to the premium the data shows is available at the upper end of the distribution.
The premium is real. It is growing. It is accessible to professionals who invest in the right capabilities rather than the most visible ones. The distance between where you are and where the premium is concentrated is measurable. Measuring it is the first step toward closing it.
Thursday we are giving you the chain-of-thought prompting framework, the technique that produces the largest single improvement in AI output quality for analytical tasks and that is, despite its outsized impact, among the least used techniques by non-technical professionals. The explanation of why it works is as useful as the prompts themselves.
— The Artificial Idea team
