Can AI Be A Solution To America’s Workforce Challenges?

Little Evidence of AI-driven Job Market Displacement

Let’s start with the evidence. The latest data does not show any evidence of widespread AI-driven job market displacement, though this is an area with lots of ongoing research. Joint research by Ramp and Revelio suggests that high-intensity AI adopting businesses are hiring more than low-intensity AI adopters. A recent working paper by Bharat Chandar and Bouke Klein Teeselink suggests that employment growth at AI adopting firms may primarily come from hiring of senior workers, in some cases at the expense of junior workers.

There is clear evidence, however, that the tasks firms expect workers to perform within jobs is changing. For instance, job postings are increasingly listing AI as a required skill. And these shifts are becoming more widespread, moving from businesses within the IT sector to many other sectors of the economy.

Asymmetric Information about Skills

Therein lies the challenge created by advances in AI, but therein also lies the opportunity for AI to be part of the solution. Workers have every incentive to invest in acquiring new skills, and indeed data tracked by Revelio Labs shows a sharp increase in the number of workers investing in their own retraining by obtaining AI certifications. But workers also lack the same information that businesses have about which skills are important now, and which skills will be important in the future. The term economists use for this is asymmetric information.

This isn’t a good outcome for anyone. It is obviously a bad outcome for the worker, especially if learning the new skill took time and money. But it is also a bad outcome for businesses. It potentially means it will take them longer to find, if at all, someone with the skills they need.

The Need for AI-enabled Skills Training

A publicly-funded AI-enabled skills training program can potentially solve this challenge on a broad scale. Such tools would identify the new opportunities for workers based on their individual background and then create a personalized training program. There’s already lots of evidence that generative AI boosts productivity for inexperienced workers, including in software coding, writing and customer support services.

There also already exist successful corporate retraining programs along these lines. For example, IBM’s “Your Learning System” uses AI to evaluate employees’ existing skills, recommend individualized pathways for skill development, and help design and update training programs. Local workforce development leaders, aided with federal funding, could begin to pilot programs through community colleges or other providers to see what works well and what doesn’t outside of the corporate environment. The National Applied AI Consortium, a new National Science Foundation funded consortium of three large community college systems and several large technology firms, potentially provides a blueprint for how to develop curriculum and train faculty to provide AI-enabled skills.

The current status quo, in which workers are constantly trying to predict which skills businesses will require, is likely contributing to the significant worries that workers have about the effects of AI on job opportunities. Such concern has in turn led to policy proposals explicitly designed to stop or slow the adoption of AI, such as a tax on AI tokens or a “robot tax” on businesses adopting AI. While such policies are aimed at protecting workers, they don’t help workers learn the skills needed to remain relevant, won’t help the U.S. remain a leader in AI, and will slow the diffusion of benefits from AI. On the other hand, an AI-enabled skills training policy could be designed to help all workers so that their skills remain relevant as work changes.

Finally, investment in labor policies, such as AI-enabled skills training programs, should go hand in hand with renewed investment in a robust federal statistical infrastructure. Good policy requires good data. Without good data, policymakers won’t be able to accurately assess the efficacy of AI-enabled skills training or other programs.