Jenuel Oras Ganawed

Posted on Oct 3 Originally published at blog.jenuel.dev

AI Made Me Faster. Shouldn't I Be Paid More?

#ai #programming #productivity #career

If one employee can now do the work of two, who keeps the difference?

AI has changed the speed of my work.

I can move from an idea to research, a draft, code, testing, and something ready to publish much faster than I could before. Tasks that used to consume a whole afternoon can sometimes be finished before lunch. The obvious thought is: if I can produce more value in the same number of hours, shouldn't my salary increase?

But there is another possibility that is harder to ignore.

If everyone can produce more with AI, perhaps the work becomes cheaper. Maybe employers stop paying a premium for tasks that used to require years of experience. Maybe one employee is expected to carry the workload of two people without receiving two salaries. Maybe the company captures the productivity gain while the worker receives a busier day.

So which is it? Does AI push salaries up because workers become more productive, or push them down because human work becomes easier to replace?

The honest answer is: both are happening, but not to the same people.

AI does not automatically raise salaries. It changes the bargaining position of workers. Pay rises when AI makes a person's judgment, experience, or specialized skills more valuable. Pay can stagnate or fall when AI makes the person's output easier for someone else to reproduce.

That difference matters more than the number of tasks completed.

A salary is not a productivity score

I used to think about salary in a fairly simple way: become better, finish more work, and eventually earn more. That relationship exists, but it is not automatic.

A salary is the price an employer must pay to keep or replace a worker. Productivity matters, but so do scarcity, bargaining power, demand, competition, ownership, and how easily the work can be measured.

Suppose AI lets me complete twice as many reports. Several outcomes are possible:

The company gives me a raise because my output became more valuable.
The company keeps my salary unchanged and enjoys a larger profit margin.
The company raises my workload until the saved time disappears.
The company needs fewer people doing the same work.
The price of the service falls because every competitor can now produce it faster.
I use the extra capacity to create my own product, accept more clients, or move into a better-paid role.

The technology creates the productivity gain. The employment relationship decides who receives it.

Historical evidence suggests that higher economy-wide productivity usually contributes to higher wages over time, including for workers near the bottom of the wage distribution. But those gains are not distributed equally. Research on the digital revolution found that the highest earners generally captured the largest wage increases as technology became more valuable.[7]

That is an important warning for the AI era. A rising productivity number does not mean everyone gets an equal raise.

We already know AI can increase output

The productivity evidence is no longer based only on demos.

A study of 5,179 customer-support agents found that access to an AI assistant increased issues resolved per hour by 14% on average. Novice and lower-skilled workers improved by 34%, while experienced workers gained much less.[11]

In another experiment involving 453 professionals, ChatGPT reduced the time required for writing tasks by 40% and increased independently rated quality by 18%.[12]

Companies expect those gains to spread. A 2026 NBER survey of nearly 750 corporate executives found that more than half of their firms had already invested in AI. Executives expected AI-related labor-productivity gains to strengthen, especially in finance and high-skill services, although perceived improvements were often larger than measured improvements.[16]

A separate survey of nearly 6,000 executives in the United States, United Kingdom, Germany, and Australia found that nine out of ten reported no measurable effect from AI on employment or productivity during the previous three years. Yet the same executives expected AI over the next three years to raise productivity by 1.4%, raise output by 0.8%, and reduce employment by 0.7%.[18]

That gap between today's measurements and tomorrow's expectations explains some of the confusion. AI is visibly changing individual workflows, but changes in company revenue, staffing, and salaries take longer to appear.

Recent firm-level research provides some evidence that the gains are beginning to show up. Companies investing more heavily in AI experienced faster productivity growth from 2018 to 2024, particularly when they used AI-skilled employees to build durable company knowledge and processes. Employment did not fall overall in that study, high-skilled employment increased, and average wages rose.[17]

That sounds hopeful. But a firm's average wage can rise even if it hires more expensive specialists while ordinary employees receive no raise. We have to distinguish a company becoming more productive from every worker sharing the benefit.

The first surprise: AI users are not automatically earning more

One of the strongest early studies on actual earnings comes from Denmark. Researchers connected surveys about chatbot adoption with administrative employment records covering roughly 25,000 workers and 7,000 workplaces.

Workers reported productivity benefits. Employers introduced new AI-related responsibilities. Work was reorganized around content generation, AI oversight, and integration. Yet two years after ChatGPT's release, the researchers found no detectable average effect on earnings or recorded hours, even among frequent users and workers who said AI saved them time. Their estimates ruled out average effects larger than about 2%.[4]

That result feels familiar. A tool can make someone's day easier or faster without changing the salary attached to the job title.

An employer usually does not say, "You produced 20% more this month, so we will immediately increase your salary by 20%." Salaries are reviewed periodically. Budgets are fixed. Individual output can be difficult to isolate. Sometimes the employee does not tell anyone how much time the tool saved. Sometimes management simply turns the faster workflow into the new baseline.

Another NBER study offers an uncomfortable possibility. Workers in occupations that complemented AI worked approximately 2.75 additional hours per week. The study also found positive associations between AI complementarity and hourly wages, but employee satisfaction declined. The researchers argue that when AI raises the value of another hour of work, employers and workers may extend the workday instead of converting all the gain into leisure.[6]

In other words, AI can raise pay and still leave a person worse off if expectations and working hours rise faster.

The second surprise: AI skills really do carry a premium

If ordinary users are not automatically receiving raises, why do we keep seeing headlines about enormous AI salaries?

Because employers are paying more for scarce AI-related skills.

PwC's 2026 Global AI Jobs Barometer analyzed more than one billion job advertisements across 27 countries and territories. It reported an average 62% wage premium for jobs requiring AI skills, up from 57% a year earlier. Jobs asking for specific AI skills were also growing much faster than the wider job market.[1]

An Oxford study of more than 10 million UK job vacancies found a smaller but still substantial result. AI skills were associated with a 23% wage premium across roles and a 36% premium in science, engineering, and technology jobs. The AI premium exceeded the study's estimated premium for a master's degree, although it remained below the premium for a PhD.[2]

Those numbers sound like proof that AI raises salaries, but there is an important limitation: job advertisements are not the same as a raise for an existing employee.

A role requiring machine learning, data engineering, model evaluation, or AI product ownership may already be more complex and senior than the average role. Employers may advertise a high salary because they need a scarce combination of experience and AI knowledge. The posting does not prove that adding ChatGPT to an ordinary workflow increases the salary by 62%.

Actual payroll data produces a more modest picture. Ravio's 2026 European technology compensation report found 88% year-over-year growth in AI and machine-learning hiring, with an average pay premium around 12%, varying by level. That is still meaningful, but it is far below the most dramatic advertised-wage headlines.[3]

My reading is that an AI premium exists, but employers pay for more than tool usage. They pay for people who can connect AI to expensive business problems, evaluate its failures, build reliable systems, and take responsibility for the result.

Knowing how to open an AI chat window will not remain scarce. Knowing what to ask, what to reject, and how to turn the output into dependable value can remain scarce for much longer.

When AI can lower pay

The same tool that increases my output can reduce the market price of that output.

Imagine that a simple website once required a week of developer time. AI reduces the work to one day. A developer may complete more projects and earn more. But competitors can do the same. More people can enter the market. Clients learn that the job is faster. The price may fall from a week's fee to a day's fee.

Productivity increased. The worker's income did not necessarily increase with it.

Online freelance markets provide early evidence of this effect. After ChatGPT's release, postings for automation-prone writing and coding work fell by 21% relative to manual-intensive work. Competition among freelancers increased. The jobs that remained tended to be more complex and better paid, but there were fewer routine opportunities available.[10]

This creates a barbell-shaped market. Basic work becomes cheaper or disappears. Difficult work still pays well, sometimes better than before. The middle can become uncomfortable.

A 2026 Apollo white paper offers a more pessimistic finding. Using occupation-level wage data and observed AI-usage measures, it estimated that real wage growth in highly exposed U.S. occupations was 6.7 percentage points lower after 2023, with no detectable employment effect. The negative estimate was larger among lower-paid workers.[5]

That result deserves caution. It is an early, non-peer-reviewed analysis built from occupation-level data, and the period includes many economic changes besides AI. It does not prove that AI caused every difference it measured. But it raises a plausible scenario: companies may keep roughly the same number of employees while allowing pay growth to slow because each worker can produce more.

Pay can fall without a dramatic layoff announcement. It can happen through smaller raises, weaker freelance rates, fewer promotions, reduced hiring, or higher expectations for the same salary.

Entry-level workers may feel the pressure first

AI is especially good at the work companies traditionally give to beginners: first drafts, simple analysis, basic code, routine customer responses, documentation, and information gathering.

Stanford researchers examining payroll records for millions of U.S. workers found no evidence of broad economy-wide displacement. They did, however, find that employment for workers aged 22 to 25 in highly AI-exposed occupations was 19% below the path implied by less-exposed peers. The gap came mainly from reduced hiring rather than more workers being fired. Base pay showed less adjustment than employment.[9]

This may be one reason existing employees do not immediately see their salaries fall. Companies can change the workforce gradually by replacing fewer departing workers and hiring fewer juniors.

For someone already established, AI may increase leverage. For someone trying to enter the profession, AI may remove the small tasks that once served as training and proof of ability.

Even so, the U.S. Bureau of Labor Statistics still projects employment for software developers, quality-assurance analysts, and testers to grow 10% from 2025 to 2035, much faster than the average occupation. It expects strong demand from AI, robotics, automation, security, and the continuing expansion of software products.[15]

The job is not simply disappearing. The entry requirements and valuable parts of the job are changing.

The key distinction: does AI complement me or commoditize me?

I think this is the most useful question a worker can ask.

AI complements me when it makes my judgment, relationships, domain knowledge, or responsibility more productive. A doctor who uses AI but remains accountable for the diagnosis may handle information better. A senior developer may use an agent to implement code while focusing on architecture, security, and product decisions. A marketer may generate drafts quickly but still own the strategy and customer understanding.

AI commoditizes me when the buyer mainly wants an output that the tool can now produce cheaply and consistently. If the work is easy to specify, easy to verify, and available from thousands of people using the same models, the price will face downward pressure.

Research on firm-level AI exposure supports this distinction. Tasks with greater AI exposure experienced lower labor demand, but productivity gains at AI-adopting firms increased demand elsewhere. Overall employment effects remained modest because substitution in some tasks was offset by growth and reallocation in others.[8]

The IMF's analysis of observed AI usage across more than 100 countries found that current gains are concentrated in higher-wage professional occupations. The concentration is particularly strong in lower-income countries, where AI use has not spread as widely across the rest of the workforce.[14]

AI may narrow performance gaps inside a task while widening economic gaps between people who control valuable systems and people who supply easily replicated output.

OECD research found no clear evidence that AI exposure changed wage inequality between occupations during the 2014–2018 period, but it found some evidence of lower wage inequality within exposed occupations. One possible explanation is that lower-performing workers gain more from AI than top performers.[13]

That sounds fairer, but it can have a darker interpretation. If AI makes average workers perform more like experts, employers may become less willing to pay for routine expertise. The premium moves toward the people who design the workflow, own the client relationship, carry legal responsibility, or solve problems the model cannot handle.

Who captures the gain?

Suppose I use AI to save ten hours each week. Who owns those ten hours?

If I am paid a fixed salary, my employer may own them. If I freelance and charge by the hour, the saved time may reduce my bill unless I change to value-based pricing. If I charge per project, I may keep the gain by completing more projects. If I own the product, the productivity gain can become profit or growth. If I use the time to learn a scarcer skill, it may become a future salary increase.

This is why the same technology creates very different outcomes for employees, contractors, and business owners.

The IMF warns that AI could increase wealth inequality even in scenarios where wage inequality narrows. Owners of AI systems, companies, and capital can receive higher returns, while workers depend mainly on what happens to their wages. High-income workers are also more likely to own assets and to hold jobs where AI complements rather than replaces their contribution.[19]

The question is not only whether AI creates value. It clearly can. The harder question is whether the value appears in wages, profits, lower prices, shorter working hours, or returns to the owners of capital.

Markets and company policies make that choice. AI does not.

What I would do as an employee

I would not walk into a salary discussion and say, "I use AI now, so I deserve more money." AI usage by itself is becoming ordinary.

I would show the business result.

How much time did the workflow save?
Did output increase without increasing errors?
Did I help the team avoid hiring a contractor?
Did I shorten delivery time?
Did I improve revenue, reliability, customer retention, or support capacity?
Did I build a repeatable system that other employees now use?
Am I responsible for checking the AI's work and handling the difficult exceptions?

A stronger salary argument sounds like this:

I redesigned this process, reduced delivery time from five days to two, maintained quality, documented the workflow, and helped the rest of the team adopt it.

That is harder to dismiss than saying I wrote more prompts.

I would also avoid giving away every productivity gain invisibly. If I quietly use AI to finish twice as much work, management may only see that the new workload is normal. Tracking before-and-after measurements makes the improvement visible.

Most importantly, I would move toward ownership. That can mean owning a system, a customer relationship, a product area, an architectural decision, a compliance risk, or a measurable business outcome. The closer my contribution is to an expensive consequence, the harder it is to price me as generic output.

What employers should do

If a company receives all the value from AI while workers receive only larger workloads, employees will learn to hide efficiency gains or disengage from the process.

Companies should consider sharing the gain through some combination of:

Higher salaries for expanded responsibility
Performance bonuses tied to measurable outcomes
Profit sharing or equity
Shorter workweeks or protected focus time
Training and promotion pathways
Clear policies about how productivity data will affect staffing

This is not only about fairness. Workers are more likely to adopt AI honestly when they do not believe the tool will immediately be used against them.

Employers also need to preserve entry-level learning. If AI performs every junior task, the company may enjoy lower costs today and discover later that it has no experienced employees ready to take responsibility.

So, will salaries rise or fall?

My answer is not satisfying, but I think it is the truthful one.

Average wages may rise in the long run if AI creates new products, expands demand, and raises economy-wide productivity. Historical evidence supports that possibility. But individual workers should not assume that doing more tasks automatically leads to a raise.

Salaries are more likely to increase when:

AI skills remain scarce.
AI complements domain expertise rather than replacing it.
The worker owns decisions and outcomes, not only production.
The company can connect the worker's contribution to revenue or avoided cost.
The worker can negotiate, change employers, freelance, or build a product.

Salaries are more likely to stagnate or decrease when:

AI makes the output easy for many people to reproduce.
Employers can measure higher output but not individual judgment.
Entry-level supply increases while routine demand falls.
Workers have little bargaining power.
The productivity gain is captured through profits, lower prices, or higher workloads.

The sentence I keep coming back to is this:

AI does not pay people for doing more. The market pays people for being difficult to replace.

That may sound harsh, but it also points toward a practical strategy. I do not need to compete with AI at typing, drafting, searching, or generating routine code. I need to use it to move closer to the parts of work that still carry responsibility: choosing the right problem, understanding the customer, judging risk, integrating systems, and owning the result.

AI can make me faster. Whether it makes me better paid depends on what I do with the speed, who can copy my output, and whether I have enough leverage to claim part of the value I create.

Sources

Originally published at https://blog.jenuel.dev/blog/ai-made-me-faster-shouldnt-i-be-paid-more

Thanks for reading! If you enjoyed this article and like this kind of content, you're always welcome to buy me a little coffee, but only if you'd like to. No pressure at all, and either way I'm truly grateful you stopped by. ☕️

DEV Community