Skip to content
Workforce 2100Research on AI, work and education
What we study Research Surveys About Contact Explore the data

Surveys · Research study

In development: opens 2027For presidents and provosts

Tracking AI's Effect on College Programs: A Multi-Year Study

The Work 2035 Tracking Study. Recruitment opens after independent ethics review, expected spring 2027.

Last updated November 18, 2026

SummaryWhy this studyWhat the study will doTimelineTaking partHow data are handledEthics reviewOpen scienceQuestions institutions askHow to get involved

Summary

The Work 2035 tracking study checks, every year and in public, how well the Work 2035 model's forecasts match what actually happens. For every college and university in the country, it publishes a yearly comparison of what the model expected with what happened. For a group of 30 to 50 participating institutions, followed over at least five years, it also asks leaders what they changed and why. The goal is to learn how well AI's effect on work can be forecast, and which college responses work.

Why this study

Colleges are making program, budget and hiring decisions about AI with little evidence. The Work 2035 model, published openly at workforce2100.org, estimates how demand for work may shift through 2035 for every occupation, region and college program in the country. According to the model and today's data, AI will reshape far more programs than it eliminates, and the effects differ sharply from one region to the next.

Those are forecasts, and forecasts need checking. Most AI-and-jobs studies are published once and never tested. This study tests the model in public, year by year, and records how institutions respond. The results will improve the model and give colleges evidence they can use.

What the study will do

Research questions

  1. How closely do the model's forecasts match what happens in each institution's programs and local job market?
  2. How are colleges and universities responding to AI: redesigning curricula, opening or closing programs, serving displaced adult workers?
  3. Which responses are associated with better outcomes for graduates and regions?

Data

  • Public data, every year: program completions reported to the federal government (IPEDS), regional employment and wage data, and other public labor-market indicators. These need nothing from any institution.
  • A public yearly comparison for every institution: what the model expected for its programs and local job market, set beside what happened. It is added to each free college profile.
  • A short annual survey of participating institutions' leaders (about 15 minutes) on the decisions they made and why.
  • An optional interview (about 45 minutes) with a small number of participating institutions each year.

Timeline

WhenStep
Winter 2026–27Conversations with associations and institutions; ethics (IRB) review
Spring 2027Invite institutions, after review approval
Fall 2027Baseline survey of participating institutions
Each fall afterUpdated forecasts; public yearly comparison added to every profile; annual survey of participants
Each spring afterCombined findings published openly

Taking part

Who can join. Accredited US community colleges, universities and university systems. We aim for a mix of sizes, regions and missions, including institutions that serve many adult and working learners.

What it involves

  • One short survey a year, completed by the president, provost or someone they name.
  • An optional interview in some years.
  • No internal data, unless an institution chooses to share it.
  • Institutions can leave the study at any time, without giving a reason.

What participants receive, at no cost

  • A personal walkthrough each year of the institution's profile and its yearly comparison.
  • How the institution's comparison sits beside those of similar participating institutions, shown without naming them.
  • Early access to the combined findings before publication.
  • A chance to shape what the next edition of the model measures, including the effects of agentic AI.

The profiles and yearly comparisons themselves are free and public for every institution, whether or not it takes part.

How data are handled

  • Public data stays public. Profiles and yearly comparisons are built from public data, cover every institution and are published openly under each institution's name.
  • What participants tell us stays confidential. Survey and interview responses are stored securely. Findings that draw on them are reported in aggregate, and an institution is named only with its written permission.
  • Used only for this study. Participants' contact details and responses are never shared with any company and never used for marketing.
  • Corrections. Any institution can report an error in its profile or comparison at athalass@gmail.com, and every report gets a reply.

Ethics review

The study will be reviewed by an independent Institutional Review Board (IRB) before any survey or interview data are collected. No institution will be asked to take part until that review is approved. Once approved, the board's name, the approval reference and a contact for participants' rights will be listed here.

The public profiles and yearly comparisons use only public data about institutions, not information about people, so they do not require this review.

Open science

The Work 2035 model, its methods and the study's combined findings are published openly at workforce2100.org under a Creative Commons Attribution (CC BY) license. The study, data and code are registered on the Open Science Framework (osf.io/ae95y). The model is re-run every year, and local industry data are refreshed quarterly.

Questions institutions ask

Is anything for sale?
No. The profiles, yearly comparisons, data and findings are free and openly licensed.
Will our survey answers be made public?
Not under your name without written permission. Published findings that use survey or interview answers are reported in aggregate.
Our profile already appears on the site. Do we have to take part?
No. Every institution's profile and yearly comparison are published from public data either way. Taking part adds your leaders' perspective and the benefits listed above.
What if the forecast for us looks wrong?
It may be. The forecasts are model estimates based on today's data, not verdicts or advice, and checking them is the point of the study. Tell us, and we will review it and reply.
Do you need our internal data?
No. Share it only if you want to.
How much time does it take?
About 15 minutes a year for the survey, plus an optional interview.
Can we leave?
Yes, at any time, without giving a reason.

How to get involved

Recruitment opens after ethics review, expected in spring 2027. To hear when it opens, or to ask a question, email athalass@gmail.com. Associations and researchers interested in collaborating are welcome to write too.

Contact
Angelo Thalassinidis, Ph.D. · athalass@gmail.com · workforce2100.org · ORCID 0009-0006-1057-0738

← Back to Surveys

Workforce 2100 · Text and our own data under CC BY 4.0, except third-party data as credited
Contact
LegalPrivacyTerms
AboutFAQSurveysResearchExplore the data