Skip to content

When Classification Fails, Compensation Follows: An AI-Assisted Job Classification Framework for Large Public Sector Employers

Sep 2026 · Compensation & Benefits Review · 0 citations · 8 references
Ethics and Social Impacts of AI

Abstract

This case study describes a classification reform at a large urban U.S. public school district of 11,480 employees and $724 million in payroll. An internal pay equity review found pay gaps between demographic groups that job content and tenure did not explain. The gaps came from how jobs had been assigned to titles and grades over the years, not from the pay decisions supervisors made within a role: single titles spanned 15 or more pay grades, and salary ranges within one role exceeded a 4.71-fold spread. The reform paired PRISM, an AI-assisted framework that classified 992 job descriptions through five validated passes, with PERCS, a model that placed each employee on the corrected grades. Estimated annual cost ran from $3 million in savings under initial placement to roughly break-even once every employee was protected against a pay cut. Human judgment governed every decision; AI informed it but never replaced it.

View source

Similar papers

Review Open access Sep 2026

Who Controls the Gains from Workplace AI? Time, Targets and Pay

Workplace AI changes the use of paid time before it necessarily changes workers' pay. This commentary synthesizes recent experiments and worker surveys alongside earlier implementation cases to explain how efficiency gains reach employees. Three allocation patterns emerge. In a Danish study, 85% of chatbot users report...

Natalie Goh, Dilorom Usmonova, N. Qodirova et al. · 0 citations
Review Sep 2026

Classification Drift: Decisions Passing Quietly but Compounding Inequity

Public-sector classification systems accumulate classification drift, a widening gap between what employees actually do, how their work is described, and how they are compensated. Because classification shapes pay, role comparison, and internal equity at scale, classification drift quietly compounds into measurable ine...

W. Birch, Martin Renfro · 0 citations

Who Benefits from Salary Prediction Tools? A Simulation Study of Worker Power and Wage Inequality in AI-Mediated Salary Negotiations

A labor market model where firms and workers are simultaneously learning to negotiate over time under different forms of pay transparency, including the presence of a salary prediction tool is developed and it is found that salary prediction tools can introduce more diverse information in the market by encouraging work...

Serafina Kamp, Kathleen A. Creel, Sina Fazelpour · 0 citations
Open access Oct 2026

More Monitoring, Less Pay: Active Ownership and Employee Compensation

The rise of active owners—hedge funds and private equity firms—has raised questions about the influence of corporate owners on employee compensation. Prior research documents that active owners reduce overall compensation through workforce restructuring and slower wage growth. What remains unexplored is whether active...

C. Gartenberg, Elaine Pak · 0 citations
Open access Aug 2026

Does it Pay to get a CPA? How License and Skill Levels Impact Accountant Salary

The United States has a severe shortage of qualified accountants. This crisis did not arise overnight; moreover, convincing data is needed to motivate stakeholders to make necessary changes. One motivating factor is to showcase how professional license and an increase in accounting skills could increase employee compen...

Katie Landgraf, Katelynn Hopson, Amy Cooper · 0 citations
Open access Sep 2026

Do Relative Wages Matter for Public-Sector Retention?

We study how relative wages are associated with sectoral exit across public-sector branches in the Netherlands. Sectoral exit is defined as no longer being employed in the same public-sector branch one year later. Using administrative individual-job data from Statistics Netherlands covering 2013–2022, we apply one-to-o...

A. Rutten, Justus van Kesteren · 0 citations

Related blog posts

MIT News · Artificial Intelligence Oct 23, 2026

Q&A: How AI came to model the “messy” human mind

Assistant Professor Benjamin Lindquist studies how engineers deliberately injected randomness into early AI, giving rise to deep-learning models that sometimes hallucinate.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.