How might artificial intelligence affect Texas’ good jobs?

The majority of research on artificial intelligence (AI) and its effect on work estimates that AI will have the greatest effect on college-educated workers and those with high incomes. Relatedly, there has been intense media attention on AI and its supposed negative effect on the job market for recent college graduates.

While AI will certainly affect white collar jobs to some degree, recent Dallas Fed research emphasizes that AI can either help or replace workers depending on the nuance of their occupations. But how might AI affect so-called good jobs, those that don’t require bachelor’s degrees but still pay self-sustaining wages?[1]

For people without four-year degrees, good jobs can be a path toward economic mobility and financial security. For this reason, these jobs are often targeted occupations of job training and career readiness programs. How should those seeking these good jobs think about AI and ready themselves for any changes to these jobs?

Comparing these jobs with the skills and work activities most exposed to AI can identify which jobs might see the most change. Jobs that involve writing, gathering information and computer programming are more exposed to AI change, while jobs that involve operating machinery or jobs in the trades tend to be less exposed.

It's important to emphasize that AI-prompted change can be positive or negative for the worker, depending on how the change impacts the job. And one set of skills seems so far impervious to AI: soft skills, which are important for all good jobs.

How much are Texas’ good jobs exposed to AI?

The top ten good jobs in Texas employed more than 1 million workers in 2023 (Table 1). We use the MIT Living Wage calculator estimate for a living wage for one adult with no children in Texas, which is $21.82 per hour.

Importantly, an occupation’s exposure to AI can be either positive (enhancing a person’s ability to do the job) or negative (reducing the need for a human to perform the job altogether). Without making judgment calls on whether exposure will be positive or negative, a review of AI’s capabilities can shine a light on how much change to expect for these jobs.

To do this, we use information from the Department of Labor’s O*NET database, specifically the skills humans need to do each good job and the work activities associated with the job.[2] We give each occupation two scores: a Skill Exposure Score, measuring how exposed the worker’s skills are to AI, and a Work Activities Exposure Score, measuring how exposed the job’s activities are to AI.[3]

Certain skills and work activities expose good jobs to AI

Our results are shown on a spectrum of less exposed to highly exposed, using two more occupations as examples of jobs at the two extremes. Locker room, coatroom and dressing room attendants represent occupations with relatively little exposure to AI, while computer programmers represent those with relatively high exposure to AI.

A couple of occupations seem to be highly exposed to AI in terms of both skills and work activities. The first is first-line supervisors of office and administrative support workers, which encompass job titles such as accounting managers, accounts payable supervisors and customer service managers, among others. The second is computer user support specialist, which includes job titles such as technical support specialist, help desk analyst and information technology specialist.

These two occupations share a number of skills and work activities that AI can perform at a high level and therefore contribute to their high exposure.

Other occupations consistently score relatively low when it comes to AI exposure. Heavy and tractor-trailer truck drivers, automotive service technicians and mechanics and electricians all have fairly low scores in both dimensions. The skills and work activities that AI cannot currently perform well and that contribute to their low scores include:

The remaining good jobs are mixed when it comes to exposure to AI.

Soft skills not highly exposed to AI

Our results clearly show that jobs requiring some kind of physical activity, whether it’s repair, installation or operation, are not very exposed to AI at this point. This kind of work may be relatively unchanged in the short- to medium term.

But again, exposure to AI is neither inherently positive nor negative. Jobs with high AI exposure scores still require skills and work activities that AI cannot do well at this point, including active listening and establishing and maintaining interpersonal relationships.

Since humans are still much better at soft skills than AI, we estimate these skills will be highly valuable in the short- to medium term.

There can be room for humans in both low- and high-exposure jobs

Good jobs that require less than bachelor’s degrees but pay living wages are an important part of the conversation on AI and the future of work. Even occupations that are less exposed to AI will still be affected to some degree. Workers without four-year degrees searching for better job opportunities would benefit from making informed choices on which jobs can give them living wages and how AI might impact these jobs going forward.

For young adults at the beginning of their careers and those that guide them through this period in their lives, there are two major takeaways from this study.

• Good jobs that work with machinery and in the trades appear to have limited exposure to AI-led changes.
• Soft skills are important for all good jobs.

For those working in high-exposure occupations, honing their soft skills may be an effective strategy to either offset negative AI exposure or complement the technical tasks AI may take off their plates.

Notes

Special thanks to Amy Higgins, Ruby Martinez-Berrier and Xiaohan Zhang for their assistance.

1. Good jobs in other contexts often include other elements such as benefits and predictable scheduling. In this study, we define “good jobs” as simply those that pay living wages and require less than bachelor’s degrees, not taking other elements into consideration.
2. “Sales representatives of services, except advertising, insurance, financial services and travel” was in the original top ten, but there was not enough data in the O*NET database on this occupation to include it in our study. This occupation was replaced by “First-line supervisors of production and operating workers.”
3. These scores are comprised of the level and importance of the job’s top skills and work activities as well as an AI assistance category (low, medium or high) and corresponding AI assistance score. The AI assistance categories and scores were primarily created by humans with input from an AI platform. For more details on methodology, please see the appendix.