AI Gives Workers a Second Chance
-
October 16, 2026
Downloads
Download Article
-
Consider a 35-year-old warehouse supervisor who did poorly in high-school, entered the army at 18, received a GED and joined a company in a role that required few formal qualifications. Even after 15 years at that company, to the outside world, he may still be a man with a GED. To his employer, however, he is someone who can solve problems, deal with difficult people, show up when needed, exercise judgment and take on responsibility. His employer knows that he can manage a team, stay calm when things go wrong and learn from mistakes. A high school record may have described this person at 18. At 35, it says little.
A job is not a collection of skills, but a bundle of tasks that change over time. An accountant runs numbers and decides which numbers to distrust. A programmer writes code in multiple languages and knows which bugs will matter in production. Because a job involves such a wide variety of tasks and judgments, worker performance is determined by far more than one’s education, technical knowledge and degrees.
Yet one’s early education still determines which jobs are available to them. Traditionally, a worker could possess most of the skills required for a job and still be excluded because they lacked specific knowledge or credentials. Employers had little choice but to demand the entire package of skills associated with a role.
AI helps bridge this gap, by enabling high-performing but less credentialed employees to undertake tasks that were previously beyond their reach. It also helps employees enter new roles without possessing the credentials traditionally associated with them. As a result, AI raises the value of individual judgment, character and experience, and makes a company’s knowledge of its own employees even more important.
The Importance of Work Performance
When companies hire young workers, they do not know much about them. A degree, grades and field of study provide information about someone whose performance the employer has never observed. Once someone has been working for several years, however, a diploma is no longer the employer’s best evidence of what that person can do. Years of actual performance are.
Research backs this notion. A 2016 study using personnel records from two large companies suggests that educational credentials become less influential in assessments of employees’ abilities, the longer their work is observed. Specifically, the authors estimate that expectation errors around an employee’s abilities fall by half within three to six years.1
The “Second Chance” Workforce
The labor market is unforgiving of what people did when they were young. Companies, however, routinely discover that people change. A decade before ChatGPT was released to the public, for example, Amazon established its Career Choice program to help hourly employees qualify for high-demand occupations. More than 8,900 employees have since earned GEDs and more than 2,000 have obtained high school diplomas through the program.2 At Starbucks, more than 5,500 employees who did not initially qualify for admission to Arizona State University subsequently entered through the company’s Pathway to Admission program.3
There is nothing particularly unusual about companies subsidizing education; roughly half of U.S. employers do so.4 What is interesting here is what happened to the workers. They entered these companies with educational records that suggested one thing about their abilities, while years of work revealed something else.
Until now, acting on that new information often required sending a worker back through the educational system. Now, AI may provide another route.
AI Changes What a Worker Needs to Know
In 2024, researchers studied the introduction of a generative AI assistant among more than 5,000 customer-service workers. Productivity increased on average for all, but the largest gains occurred among less experienced and lower-skilled employees. The AI system presumably helped lower-skilled workers adopt some of the practices used by the organization’s better performers.5
More recent experimental evidence is striking. In a randomized controlled experiment, 1,174 adults were given a workplace-style problem-solving task with or without access to generative AI.6 Without AI, participants with more education outperformed participants with less education by 0.55 standard deviations. With AI, the gap was 0.14, representing a 75% reduction in the initial education-related productivity gap.7
AI changes what workers themselves need to know. While AI can supply information, perform calculations, produce a first draft, explain an unfamiliar concept or guide someone through a task, the worker must still know what to ask, recognize when the answer makes sense and decide what to do with it.
But AI will not erase the value of education. Nor will someone without specialized training suddenly be able to work in any job. Some work genuinely requires knowledge, training or licensing that AI cannot replace. Rather, AI shifts the question from whether someone has the traditional credentials for a job to which parts of those credentials are necessary to do the work.
It’s Time to Re-Examine Employees
Companies often discuss AI as a tool to make their existing workforce more productive. But they should also use it to re-examine employees who have been screened out of better jobs.
Instead of asking whether an employee is qualified for a job as it has traditionally been defined, they should ask which parts of that job they can do, which parts AI can help them perform and which require additional training. There are obvious limits, of course. Surgeons need medical training. Engineers must know whether a structure will stand. Licensing requirements do not disappear because ChatGPT exists. AI’s ability to produce plausible nonsense makes judgment more important, not less.
The opportunity lies in jobs where technical knowledge has been a barrier to advancement, even when success depends largely on judgment, reliability, experience and the ability to learn. Consider the warehouse supervisor. AI will not make him an accountant by Monday. But it can help him draft a budget memo, run a reconciliation or flag an anomaly in a spreadsheet. He must decide whether the anomaly matters, whether the assumptions make sense, and whether a number that looks accurate is accurate.
A More Forgiving Labor Market
For most of modern economic life, what people accomplished between the ages of 18 and 25 could shape the jobs available to them for the next 40 years. It’s true that there were always second chances. People returned to school, learned trades, joined the military, started businesses or simply worked their way up. But changing tracks was expensive because moving into a new occupation often meant studying an entirely new body of technical knowledge before anyone would give you the job.
Generative AI can lower that barrier. It will hurt some workers and help others, but one group that gets surprisingly little attention is people whose abilities have outgrown their credentials. A diploma tells an employer something important about what someone knew and accomplished when they were young. Watching someone work for years tells a different story about who they have become. AI does not make diplomas irrelevant – but it does make one’s years in the workforce count for more.
Footnotes:
1: Shota Araki, Daiji Kawaguchi and Yuki Onozuka, “University prestige, performance evaluation, and promotion: Estimating the employer learning model using personnel datasets,” Labour Economics, Vol. 41, pp. 135-148 (Aug. 2016).
2: “8 free skills training programs that help Amazon employees land higher-paying roles,” Amazon (Oct. 2020, updated Jun. 2025).
3: “Starbucks College Achievement Plan,” Starbucks.
4: 45% according to the SHRM. See “All Things Work Podcast, Season 3, Episode 33”, “Let’s Talk About Pay: Hard Conversations and Negotiation Strategies” Society for Human Resource Management (Feb. 2026); 57% according to the IFEBP. See “Education Benefits: 2024 Survey Report,” International Foundation of Employee Benefit Plans (2024).
5: Erik Brynjolfsson, Danielle Li and Lindsey Raymond, “Generative AI at Work,” The Quarterly Journal of Economics, Vol. 140, No. 2, pp. 889-942 (May 2025).
6: Guillermo Cruces, Diego Fernández Meijide, Sebastian Galiani, Ramiro H. Gálvez and María Lombardi, “Does Generative AI Narrow Education-Based Productivity Gaps? Evidence from a Randomized Experiment,” NBER Working Paper No. 34851 (2026).
7: Id.
The views expressed herein are those of the author(s) and not necessarily the views of FTI Consulting, Inc., its management, its subsidiaries, its affiliates, or its other professionals.
Related Insights
Related Information
Published
October 16, 2026
Key Contacts
Head of Latin America International Arbitration