Latest / Can't Find Job? AI Is Quietly Replacing Millions of Workers / AI Adoption and Labor Demand: Which Occupations Shifted?

Transcript

AI Adoption and Labor Demand: Which Occupations Shifted? · Can't Find Job? AI Is Quietly Replacing Millions of Workers

• 0:00
  SPEAKER_00: AI, adoption, and labor demand.
• 0:02
  SPEAKER_00: Which occupations shifted?
• 0:04
  SPEAKER_00: Evidence cutoff, September 24, 2026.
• 0:08
  SPEAKER_00: September is not yet a complete month.
• 0:10
  SPEAKER_00: The latest official occupation level employment figures run through August 2026.
• 0:16
  SPEAKER_00: Job posting reports also have different end dates and use different ways to count artificial intelligence skills.
• 0:22
  SPEAKER_00: The evidence below, therefore, describes the period from September 2025 through the latest available data, not a complete September to September comparison.
• 0:31
  SPEAKER_00: Introduction.
• 0:33
  SPEAKER_00: Artificial intelligence is changing what employers ask people to do.
• 0:36
  SPEAKER_00: But the evidence does not show that every exposed occupation is shrinking.
• 0:40
  SPEAKER_00: Job postings increasingly ask for artificial intelligence skills, and some software jobs have recently rebounded.
• 0:47
  SPEAKER_00: At the same time, several studies find that hiring is weaker for young workers in occupations where artificial intelligence can take over routine tasks.
• 0:55
  SPEAKER_00: A useful way to make sense of the mixed picture is to separate three kinds of change.
• 1:00
  SPEAKER_00: New demand, employers hire people to build, manage, or apply artificial intelligence systems.
• 1:06
  SPEAKER_00: Substitution, a tool performs tasks that once required a worker, so employers may need fewer workers for that work.
• 1:13
  SPEAKER_00: Complementarity, a tool helps a worker do more, while the job still needs that person's judgment, knowledge, or contact with other people.
• 1:21
  SPEAKER_00: The strongest pattern is not that artificial intelligence replaces whole occupations.
• 1:25
  SPEAKER_00: It is that demand is shifting within occupations, toward building the tools, using them well, checking their work, and handling tasks that still need human judgment.
• 1:34
  SPEAKER_00: How to read the evidence.
• 1:36
  SPEAKER_00: A job is a bundle of tasks.
• 1:38
  SPEAKER_00: The United States Department of Labor's Occupational Information Network, known as ONET, describes jobs through their tasks, skills, and work activities.
• 1:46
  SPEAKER_00: Its latest database includes 18,838 task statements and is updated regularly.
• 1:52
  SPEAKER_00: This lets researchers ask which parts of a job a tool may affect, rather than labeling an entire occupation as safe or at risk.
• 2:00
  SPEAKER_00: For example, customer service representatives may answer routine questions, keep records, and resolve complaints.
• 2:06
  SPEAKER_00: Software developers may write or change code, test programs, plan projects, and work with a team.
• 2:12
  SPEAKER_00: Registered nurses assess patients, plan care, and keep medical records.
• 2:16
  SPEAKER_00: Artificial intelligence may help with some of these tasks, but that does not mean it can do every part of the job.
• 2:22
  SPEAKER_00: Job postings are not the same as jobs.
• 2:24
  SPEAKER_00: A posting that mentions a tool or skill shows what an employer says it wants, it does not prove the tool is being used, that the role is new, or that the employer hired someone.
• 2:34
  SPEAKER_00: A posting may name a tool in passing, and one job opening may appear on several sites.
• 2:38
  SPEAKER_00: Employment statistics measure people working in an occupation.
• 2:42
  SPEAKER_00: They can show whether the number of workers changed, but they usually cannot establish why.
• 2:48
  SPEAKER_00: This distinction matters.
• 2:50
  SPEAKER_00: A rise in postings asking for artificial intelligence skills can happen even as total hiring remains slow.
• 2:57
  SPEAKER_00: What changed in job postings?
• 2:59
  SPEAKER_00: The broad trend is clear.
• 3:01
  SPEAKER_00: More employers are asking for artificial intelligence skills, but those requirements remain a minority of postings.
• 3:08
  SPEAKER_00: Lightcast data, analyzed by the Bipartisan Policy Center, show that postings requiring artificial intelligence skills rose 144% over the year through April 2026, while overall postings rose 7%.
• 3:23
  SPEAKER_00: By May 2026, postings with these skills had more than doubled from a year earlier.
• 3:28
  SPEAKER_00: The measure covers a broad set of skills, not only generative tools.
• 3:33
  SPEAKER_00: A separate lightcast analysis found 1.12 million United States job postings requiring artificial intelligence skills in 2025, a 66% increase from 2024.
• 3:45
  SPEAKER_00: They made up 2.8% of postings that year.
• 3:48
  SPEAKER_00: LinkedIn reported that United States postings requiring artificial intelligence literacy skills, such as prompt design, grew 70% year over year in its 2026 report.
• 3:59
  SPEAKER_00: These skills appeared in technical and non-technical jobs.
• 4:03
  SPEAKER_00: A September 2026 analysis by ICIMS and Lightcast estimated that artificial intelligence-related roles account for about 4% of United States hiring demand.
• 4:14
  SPEAKER_00: Its list of occupations with the highest concentration of these skills includes artificial intelligence engineers, machine learning engineers, natural language processing engineers, deep learning engineers, generative artificial intelligence engineers, and data scientists.
• 4:29
  SPEAKER_00: These figures are not directly interchangeable.
• 4:31
  SPEAKER_00: They use different skill lists, time periods, and definitions of an artificial intelligence-related posting.
• 4:37
  SPEAKER_00: Still, they point in the same direction.
• 4:40
  SPEAKER_00: Employers are adding these skills to job descriptions faster than the overall number of postings is growing, while most jobs are not artificial intelligence specialist jobs.
• 4:49
  SPEAKER_00: Demand is also spreading beyond technology companies.
• 4:52
  SPEAKER_00: The bipartisan policy center found rising demand for artificial intelligence skills in finance, higher education, engineering, and accounting.
• 5:00
  SPEAKER_00: Lightcast's earlier analysis found that more than half of postings mentioning these skills were outside traditional technology industries.
• 5:08
  SPEAKER_00: Indeed, also found that the number of United States job title categories containing AI more than tripled since 2022, with artificial intelligence increasingly appearing in non-technical titles.
• 5:20
  SPEAKER_00: Which occupations shifted?
• 5:22
  SPEAKER_00: The table separates direct hiring demand from task level pressure.
• 5:26
  SPEAKER_00: A posting trend is evidence of employer interest.
• 5:29
  SPEAKER_00: It is not, on its own, proof of net job growth or job loss.
• 5:33
  SPEAKER_00: For the full table, please open this article on can'tfindjob.com.
• 5:37
  SPEAKER_00: Indeed, software figures are a useful example of why job titles alone can mislead.
• 5:42
  SPEAKER_00: Software postings rose over the year, but much of the increase was for senior positions.
• 5:47
  SPEAKER_00: A recovery in postings, therefore, does not mean that every type of software job, especially first jobs for new workers, recovered equally.
• 5:56
  SPEAKER_00: For office and customer-facing work, the task mix matters.
• 6:00
  SPEAKER_00: A tool may answer a common question or draft a routine message, while a person handles unusual problems, explains a decision, or deals with an upset customer.
• 6:10
  SPEAKER_00: Similarly, software tools can help write or test code, while people remain responsible for choosing what to build, checking the result, and fitting it into a working system.
• 6:20
  SPEAKER_00: This is an interpretation of ONET task descriptions, not a claim that every employer has adopted these tools in the same way.
• 6:27
  SPEAKER_00: What happened to employment?
• 6:29
  SPEAKER_00: The latest available Bureau of Labor Statistics occupation table compares August 2025 with August 2026.
• 6:37
  SPEAKER_00: It counts employed people in broad occupational groups, rather than job openings or detailed occupations.
• 6:43
  SPEAKER_00: For the full table, please open this article on can'tfindjob.com.
• 6:47
  SPEAKER_00: The details inside the broad groups also moved in different directions.
• 6:51
  SPEAKER_00: Employment in office and administrative support fell by 261,000.
• 6:55
  SPEAKER_00: Sales and related work rose by 93,000.
• 6:58
  SPEAKER_00: Construction and extraction fell by 639,000, while production work rose by 234,000, and transportation and material moving work rose by 364,000.
• 7:10
  SPEAKER_00: These figures cannot be treated as artificial intelligence effects.
• 7:14
  SPEAKER_00: The table is not seasonally adjusted and counts people, not vacancies.
• 7:18
  SPEAKER_00: Its large occupational groups combine jobs with very different tasks.
• 7:22
  SPEAKER_00: The Bureau of Labor Statistics also notes that updated population controls are introduced annually.
• 7:28
  SPEAKER_00: The figures are useful as a broad check on employment, but they do not show that artificial intelligence caused a change in any of these groups.
• 7:36
  SPEAKER_00: The pattern also does not provide a simple technology jobs up, other jobs down story.
• 7:40
  SPEAKER_00: Management and professional employment rose overall, but service work fell.
• 7:45
  SPEAKER_00: Production and transportation employment rose, even though those jobs are not the main source of artificial intelligence skill postings.
• 7:51
  SPEAKER_00: The data make it especially important to separate general labor market change from changes tied to artificial intelligence adoption.
• 7:59
  SPEAKER_00: Where is substitution most visible?
• 8:01
  SPEAKER_00: The clearest warning concerns routine, repeatable tasks and the workers who often begin their careers doing them.
• 8:07
  SPEAKER_00: A September 2026 Dallas Federal Reserve study used ONET tasks and artificial intelligence use data to compare postings for jobs with different levels of task exposure.
• 8:17
  SPEAKER_00: In Texas, postings for more exposed occupations fell about 5% relative to less exposed jobs by the end of 2023, and about 8% by the first quarter of 2025.
• 8:30
  SPEAKER_00: Among existing firms, the relative decline reached about 8 to 9% by early 2026.
• 8:37
  SPEAKER_00: These estimates compare occupations with different levels of task exposure.
• 8:41
  SPEAKER_00: They do not mean that postings across the whole economy fell by those amounts.
• 8:46
  SPEAKER_00: The researchers also found that some firms shifted their postings away from more automatable tasks.
• 8:52
  SPEAKER_00: A different type of evidence points to pressure on early career hiring.
• 8:56
  SPEAKER_00: A Stanford study using payroll data through June 2026 found that employment among workers ages 22 to 25 in highly artificial intelligence exposed occupations stood 19% below the level it would have reached if it had kept pace with young workers in less exposed occupations.
• 9:15
  SPEAKER_00: The authors said the difference appeared mainly through weaker hiring, not more layoffs.
• 9:19
  SPEAKER_00: They also stressed that this is a descriptive pattern, not proof that artificial intelligence caused the full gap.
• 9:26
  SPEAKER_00: Other studies do not find the same broad signal in postings.
• 9:29
  SPEAKER_00: A New York Federal Reserve analysis found little evidence of a distinct artificial intelligence-driven drop in postings for exposed occupations.
• 9:38
  SPEAKER_00: A Federal Reserve Board study found no negative effect of its firm-level artificial intelligence adoption measures on later job posting behavior through 2025.
• 9:47
  SPEAKER_00: The evidence is therefore uneven, not settled.
• 9:50
  SPEAKER_00: One reasonable interpretation is that the earliest pressure may be easiest to see in particular tasks, occupations, and age groups, not in total national employment.
• 10:00
  SPEAKER_00: A hiring pullback can affect new entrants even when a company does not lay off current staff.
• 10:05
  SPEAKER_00: But weaker hiring can also have other causes, including slow growth in the industry, changes in interest rates, and the ordinary ups and downs of hiring.
• 10:14
  SPEAKER_00: Complementarity, where people and tools work together.
• 10:17
  SPEAKER_00: A job can be highly exposed to artificial intelligence without being a likely candidate for full replacement.
• 10:23
  SPEAKER_00: A tool may help someone complete a task faster, allowing employers to ask for more work, different work, or a different mix of skills.
• 10:31
  SPEAKER_00: This seems especially relevant for experienced software developers who can use coding tools but still need to design systems, review code, test it, and understand the business need.
• 10:42
  SPEAKER_00: Data scientists and business analysts who can use automated tools to prepare information or generate early findings while people check what the results mean.
• 10:51
  SPEAKER_00: Health workers, where tools may help with notes or information sorting, but patient care still involves human contact, judgment, and responsibility.
• 10:59
  SPEAKER_00: Engineers and manufacturing workers, where artificial intelligence and robotic skills may support design, maintenance, and production work.
• 11:08
  SPEAKER_00: Finance, accounting, and marketing professionals, where routine reporting or drafts may be automated, while analysis, compliance, client advice, and final decisions remain important.
• 11:19
  SPEAKER_00: This is a task-based reading of the evidence, not a claim that every occupation above saw net job growth because of artificial intelligence.
• 11:27
  SPEAKER_00: The Atlanta Federal Reserve's analysis of the Southeast found demand clustered in computer and mathematical work, engineering, and business operations, while most artificial intelligence postings required only a small number of such skills.
• 11:41
  SPEAKER_00: That pattern suggests employers often add these skills to existing jobs rather than create a separate artificial intelligence job for each task.
• 11:49
  SPEAKER_00: Wages and training investments.
• 11:51
  SPEAKER_00: Wage premiums depend on what is measured.
• 11:54
  SPEAKER_00: The wage evidence suggests that skills for building artificial intelligence systems may be rewarded more than basic familiarity with using them.
• 12:02
  SPEAKER_00: An International Monetary Fund study of United States job postings found a posted wage premium of more than 8% for artificial intelligence developer skills, compared with about 2% for artificial intelligence user skills.
• 12:16
  SPEAKER_00: These are posting-based comparisons, not a promise that a worker will receive that raise.
• 12:21
  SPEAKER_00: Indeed, September 2026 analysis found that advertised pay in the most artificial intelligence exposed occupations had risen about 46% since 2021, compared with 25% in the least exposed group.
• 12:36
  SPEAKER_00: After accounting for occupation mix, it estimated a 5.7% post-Chat GPT pay advantage.
• 12:42
  SPEAKER_00: Comparing the same job titles reduced the estimate to 4.7%.
• 12:47
  SPEAKER_00: Controlling for seniority reduced it further to 2.4%.
• 12:51
  SPEAKER_00: That last result matters because postings for more exposed jobs shifted towards senior roles.
• 12:57
  SPEAKER_00: The entry-level share fell from 29% to 10%, while the senior share rose from 22% to 47%.
• 13:04
  SPEAKER_00: The analysis measures advertised pay, and some of the apparent premium reflects which jobs employers are posting.
• 13:11
  SPEAKER_00: These results point to two different rewards: a higher price for specialized skills that build artificial intelligence systems, and a smaller, less certain premium for general exposure or use.
• 13:21
  SPEAKER_00: A large raw difference in advertised pay should not be read as the amount every worker will earn for adding an artificial intelligence skill.
• 13:29
  SPEAKER_00: Workers are learning faster than employers are training.
• 13:31
  SPEAKER_00: In an ICIMS survey of 1,000 United States job seekers, 47% said they had worked on their artificial intelligence skills during the past six months.
• 13:41
  SPEAKER_00: The share who said they were teaching themselves rose from 22% to 30%, while employer-provided training remained around 1 in 6.
• 13:51
  SPEAKER_00: Although 60% felt ready to adapt, 61% said their skills were mostly limited to general tools such as ChatGPT, Copilot, or Gemini.
• 14:01
  SPEAKER_00: That is a gap employers can address.
• 14:04
  SPEAKER_00: Training is more useful when it matches the job.
• 14:07
  SPEAKER_00: Customer service staff can practice checking generated answers, accountants can learn to verify automated summaries, software teams can focus on testing, security, and code review.
• 14:18
  SPEAKER_00: Employers should set clear rules about what information workers may enter into tools and give people time to practice, not simply tell them to use artificial intelligence.
• 14:28
  SPEAKER_00: Regional clusters, where demand is concentrated, artificial intelligence employment and hiring are concentrated in a handful of large technology and business centers, though postings asking for related skills appear across the country.
• 14:42
  SPEAKER_00: CBRE estimated that 37% of United States artificial intelligence jobs were in the San Francisco Bay Area, New York City, Seattle, and Washington District of Columbia.
• 14:53
  SPEAKER_00: As of June 2026, its counts included nearly 98,700 artificial intelligence workers in the Bay Area, 67,900 in the New York region, 41,600 in Seattle, 33,500 in Washington District of Columbia, and 29,800 in Dallas Fort Worth.
• 15:14
  SPEAKER_00: The Bay Area added more than 20,000 artificial intelligence jobs over the prior year, even as its overall technology workforce had contracted since 2022.
• 15:25
  SPEAKER_00: These are CBRE's market estimates, not a government count of all artificial intelligence-related postings.
• 15:31
  SPEAKER_00: Other centers matter too.
• 15:33
  SPEAKER_00: CBRE estimated about 27,500 artificial intelligence workers in Boston, 21,200 in Chicago, and 19,600 in Atlanta.
• 15:44
  SPEAKER_00: The Southeast shows how a large hub can shape a state's numbers.
• 15:48
  SPEAKER_00: In Lightcast data through December 2025, Georgia led the sixth-state region in the share of postings requiring artificial intelligence skills, at 2.93%, close to the national rate of 3%.
• 16:00
  SPEAKER_00: The Atlanta region accounted for more than four-fifths of Georgia's artificial intelligence postings.
• 16:06
  SPEAKER_00: Florida's demand was spread across several large metro areas, while Alabama's was more balanced across Huntsville, Montgomery, and Birmingham.
• 16:14
  SPEAKER_00: The practical lesson is that workers do not need to move to the Bay Area to find jobs that use artificial intelligence, but the most specialized work remains concentrated.
• 16:23
  SPEAKER_00: In many regions, the larger opportunity may be to combine a field skill, such as finance, healthcare, or engineering, with the ability to use and check artificial intelligence tools.
• 16:34
  SPEAKER_00: Practical advice for workers and employers.
• 16:37
  SPEAKER_00: For workers, learn tools used in your field, not only general-purpose chat tools, practice checking results.
• 16:44
  SPEAKER_00: Employers need people who can spot errors, protect private information, and explain why an answer is sound.
• 16:52
  SPEAKER_00: Show work, not just a course badge.
• 16:54
  SPEAKER_00: A small project or work sample can show how you used a tool to solve a real problem.
• 16:59
  SPEAKER_00: Build skills that work alongside automation, communication, problem solving, subject knowledge, and the ability to handle exceptions.
• 17:08
  SPEAKER_00: For employers, list the specific tasks a tool will support or replace before changing staffing plans.
• 17:14
  SPEAKER_00: Train workers by job role and give them paid time to practice.
• 17:18
  SPEAKER_00: Track hiring, pay, and advancement by experience level.
• 17:22
  SPEAKER_00: A company may keep its overall headcount steady while quietly reducing first job openings.
• 17:28
  SPEAKER_00: Check whether new tools improve service and quality, not only whether they reduce time or staffing.
• 17:34
  SPEAKER_00: Conclusion.
• 17:35
  SPEAKER_00: From September 2025 through the latest available 2026 data, artificial intelligence skills became more common in job postings, and demand spread beyond technology roles.
• 17:46
  SPEAKER_00: The clearest growth was in specialist jobs and in senior software work.
• 17:50
  SPEAKER_00: The clearest risk was for routine, repeatable tasks, and for some early career workers trying to enter highly exposed occupations.
• 17:57
  SPEAKER_00: But the data do not show one simple national story of jobs disappearing.
• 18:02
  SPEAKER_00: Employment counts, job postings, wages, and task exposure measure different things and studies reach different findings.
• 18:09
  SPEAKER_00: The most useful conclusion is that artificial intelligence is reshaping the task mix and the experience level employers seek.
• 18:17
  SPEAKER_00: Workers and employers should prepare for that shift while treating predictions of whole occupation replacement with caution.
• 18:24
  SPEAKER_00: All links to sources are available in the text version of this article.
• 18:28
  SPEAKER_00: You can find the full article at can'tfindjob.com slash blog.
• 18:33
  SPEAKER_00: Thanks for listening.
• 18:34
  SPEAKER_00: If today's episode hit close to home, stop scrolling job boards that weren't built for this new reality.
• 18:39
  SPEAKER_00: Check out Claw Earn on AIAagentStore.ai, the first jobs marketplace designed for both humans and AI agents, so you can start earning no matter which side of the AI revolution you're on.