AI in the Workforce: Substitution, Complementarity, and Wage Pressures

Artificial intelligence software, commonly known as AI, has rapidly proliferated over the past few years among students and professionals alike.  Popular AI programs like ChatGPT can compile many sources of online information far more rapidly than any human and produce answers that are presented in a desired format, such as common language.  Many consumers use ChatGPT to deliver easy-to-understand answers drawn from many websites at once, effectively doing hours of research in less than a minute.  Workers benefit from the same ability.

AI:  A Complement or Substitute?

On the job site, workers can use AI to compile information from their systems to deliver definitive answers.  But does this make AI a substitute for, or a complement to, the human employee?  This is a major debate due to the tremendous effects the answers could have on the labor market.  Proponents of AI typically argue that it is a complement and helps employees perform more work, making their lives easier.  Critics of AI fear that it will be used to replace employees and cause mass unemployment.

Substitute: Automation Reduces Need for Labor

If a job can be largely automated using AI, workers run the risk of being replaced.  These jobs deal heavily with data entry, translation, research, and communication, all of which are effectively done by AI programs.  Data entry jobs are considered most vulnerable, with current AI programs able to take mass amounts of data from one platform and sort them into another effectively.  This puts many bookkeeping and data entry clerk jobs at risk of being fully or almost fully automated by AI.  A large office with six bookkeepers and/or data entry clerks may be reduced to a single remaining clerk to run the AI software.

Research, programming, and translating jobs may be automated soon, with AI programs approaching the point of being able to comb libraries full of legal text and producing definitive answers far more quickly than paralegals.  Some believe that AI programs will be able to replace many computer programmers by allowing many programming functions to be automated or done by laypeople using common language.  While some programmers would still be needed for complex or creative functions, much of the repetitive work could be outsourced to AI.  In schools, businesses, and government settings, AI can currently translate text more rapidly than human translators or interpreters, though critics say it will take years for AI to process social nuances and cultural variables well enough for always-accurate audio translations. 

Some physical jobs like operating vehicles, ranging from automobiles to airliners, are seen as vulnerable to AI autopilots.  While most observers believe that human operators will not be replaced entirely, especially due to public concerns about driverless vehicles run amok, flight crews may be reduced, perhaps to a single human pilot per airliner.  Equipment operation in factories may also become more automated, with AI replacing human operators in jobs that are considered very boring - causing operator distraction and fatigue - or dangerous.

Complement: Marginal Productivity Theory

Proponents of AI tend to argue that AI will turn existing human workers into super-employees, combining the benefits of human creativity, discretion, and wisdom with the data-driven and precision skills of computing.  Thus, AI will serve as a complement to workers and increase demand for workers by making each worker more productive.  With AI, each worker will generate more marginal product (additional units of output).  This increases the marginal revenue product (MRP) of each worker, which is determined by each additional worker’s marginal product multiplied by the sale price.

Simply put, if AI helps each new worker make more stuff than before, and thereby generate more revenue than before, firms will want to hire more workers rather than less.  This will occur up until the point where the last worker’s MRP is equal to his or her marginal cost (wages plus materials used).  Assuming workers’ wages remain steady, a higher MRP per worker will lead to more workers being employed - the firm can add more workers and generate more total profit.

Fatigue Reduction Increases Hours Worked

AI may increase workers’ total pay by increasing the number of hours that can be worked.  Software and hardware that are improved by AI can help workers stay on the job longer without feeling fatigued.  For example, AI-designed equipment may be more comfortable and user-friendly, allowing workers to complete longer shifts safely.  In high demand fields, this could allow more workers to be hired in total to generate more output.

Productivity Gains and Worker Displacement

What is most likely to occur is a combination of both options.  Structural unemployment, specifically technological unemployment, will likely increase substantially in the coming years due to AI replacing relatively low-skilled office workers with automated data and communication systems.  These workers will become displaced, but many will find other jobs available due to similar AI capabilities.  For example, a data entry clerk may become technologically unemployed due to AI, but find that another AI program allows him to become an equipment operator despite minimal experience.

Wage Polarization

Unfortunately, AI may cause a growing wage divide between low-skilled and high-skilled workers (or workers whose jobs can be substituted with AI and those whose jobs cannot be substituted).  While AI may automate some jobs and open up other jobs, these jobs may be low wage due to the large number of workers who can fill them.  A risk of having AI make jobs easy is that workers become highly replaceable, keeping wages low and reducing firms’ incentives to improve working conditions.  Historically, this would be similar to what happened to workers during the 1870s and 1880s during the era of mechanization during the Gilded Age.  Output rose due to mechanization and factory production, resulting in falling prices and falling wages.  

A smaller number of workers will enjoy higher wages due to increased marginal revenue product.  Combining their skill with AI software, they can generate more output and make more income.  Artists can create more artwork, surgeons can complete more procedures, and business owners can reap the efficiency gains of AI-optimized planning and logistics.  Ultimately, the beneficiaries of AI efficiency gains are likely to be business owners, who can use AI to minimize costs by trimming lags and redundancies, increase sales through savvy marketing, and identify business opportunities even in distant markets by combing through reams of data.

A Limit:  Consumer Wariness and Government Interventions

Unlike the Gilded Age, today’s more responsive governments are likely preparing for disruptions to labor markets caused by AI.  Regulatory agencies and legislatures are likely to put forth limits on firms’ use of AI tools in the name of public safety.  For example, even if an AI program can safely pilot an airliner more effectively than most humans, laws will almost certainly require a human pilot to remain in the cockpit.  Consumers are unlikely to accept full control of heavy equipment solely by computer chips - they will want human operators to remain on the job in case of emergencies.  

Fears of mass unemployment will also influence governments to explore ways to employ those who have lost their jobs to AI, perhaps in new public sector jobs.  Additional business profits generated by AI may be taxed to allow increased government spending on infrastructure, education, and health care, thereby leading to more employment in those sectors.  Therefore, the likelihood of long-term mass unemployment due to AI is smaller than feared.  Allowing millions of workers to remain unemployed for long increases the probability of social unrest, which can lead to violence.