AI Qualitative Research: Lessons From RAND

According to Three Essays on Artificial Intelligence and Work in the U.S., AI exposure was associated with higher employment growth across higher-exposure occupations from 2005 to 2023, but the effects differed by type of work.

RAND's qualitative medical coding analysis found that AI can remove simpler tasks while increasing demand for expertise, creating both well-being benefits and training challenges for coders.

• RAND analyzed relationships between occupational AI exposure and labor market outcomes across the 2005 to 2023 period.
• RAND reported that higher AI exposure was associated with higher employment growth, without substantial changes in mean wages across the occupations examined.
• RAND found stronger employment and wage growth in information-focused occupations, while people-focused occupations experienced lower employment and wage growth associated with AI exposure.
• RAND's medical coding interviews included hospital leaders, medical coders, and AI vendors, showing why qualitative evidence is necessary for understanding implementation effects.
• RAND's dissertation was published on November 9, 2026, and recommended better measures of AI activity, task and skill data, and individual employment and wage data.

RAND's qualitative interviews show that medical coding automation can improve some aspects of coder work while weakening the traditional pathway through which new coders learn.

According to RAND's interviews with hospital leaders, medical coders, and AI vendors, AI automates simpler coding tasks that were previously performed by coders. The change can improve coder well-being by reducing routine work, but it can also remove entry-level learning opportunities.

RAND's author described the central tension this way: "AI can increase demand for expertise while complicating the ability of training systems to produce that expertise." The quotation captures a qualitative research finding that employment data alone cannot fully explain: automation can raise the value of expertise while making expertise harder to develop.

RAND's medical coding analysis also shows why researchers should code interviews for both positive and negative effects. Themes such as reduced routine burden, changing expertise requirements, weaker training pipelines, and organizational adaptation can coexist in the same AI implementation.

Researchers studying AI at work should combine labor market measures with structured qualitative evidence about tasks, skills, training, and worker well-being.

RAND's findings indicate that AI exposure does not produce one uniform employment or wage outcome. Research teams should compare occupational or organizational segments because information-focused occupations and people-focused occupations showed different relationships with AI exposure in RAND's 2005 to 2023 analysis.

RAND's medical coding interviews indicate that implementation research should ask how automation changes learning pathways, not only whether organizations reduce labor demand. Interview guides should therefore cover routine tasks, expert judgment, onboarding, supervision, morale, error management, and future skill requirements.

RAND also identified gaps in existing research data. According to the dissertation, researchers need better measures of AI activity and availability, more usable task and skill data, and more accessible individual employment and wage data.

For healthcare and policy research, qualitative findings should be treated as research evidence rather than anecdotal decoration. Medical and workplace analyses remain non-diagnostic and research-focused; interview evidence should be interpreted alongside study design, sampling, and measurement limits.

RAND's November 9, 2026 dissertation shows why AI and work research benefits from both quantitative labor market analysis and qualitative investigation of tasks, expertise, training, and well-being.

The study's 2005 to 2023 findings caution against treating AI exposure as a single, uniform force, while its medical coding interviews show how automation can improve routine work and disrupt pathways into expertise.

Research teams can use Evidano to organize large collections of interviews and documents, compare themes across segments, and make evidence synthesis more traceable. Try Evidano for free to begin analyzing qualitative research data.