AI, Jobs, and the Fight for Shared Prosperity
What university research reveals about employment, opportunity, and who captures the gains.
Rajesh Ramchandani · 2026-10-20
A parent looking at university fees and a graduate waiting for an interview are asking a more personal version of the question dominating the AI debate: Where do I fit in the future being built?
As someone building in AI, I see the opportunity to make expertise more accessible and businesses more capable. I also believe that a technology can create economic value while leaving some people with fewer opportunities. Those possibilities need to be considered together.
Recent university research gives us a useful starting point. It offers evidence of change, considerable uncertainty, and reasons to question both sweeping reassurance and predictions of inevitable mass unemployment.
The Evidence Behind the Employment Debate
On October 14, 2026, Yale’s Budget Lab updated its analysis of AI and the labor market. It reported no clear AI-related footprint in the occupational mix or its exposure-based analysis, and no connection between measures of AI usage and changes in employment or unemployment. [1]
That is a finding about the evidence available today. It is not proof that AI will have little effect in the future, or that every group of workers is doing well.
Stanford’s Digital Economy Lab presents a more specific concern. Its August 12 update, using ADP payroll data through June 2026, reports a roughly 19% employment shortfall for workers aged 22–25 in highly AI-exposed occupations relative to a benchmark based on less-exposed peers. It sees reduced hiring as the main adjustment, rather than increased separations. [2]
That figure does not mean AI eliminated 19% of young people’s jobs. The authors describe the patterns as observational and explicitly caution that they do not establish causation. They also discuss sensitivity to education, sample selection, and analytical choices. [2]
My reading is that these results can coexist. A broad labor market may show no clear overall displacement while particular groups face a harder path into work. The studies use different approaches; they should not be presented as a settled contest with one winner.
When the First Rung of the Career Ladder Disappears
For a family, the difference between a lost job and a job that never opens can be academic. Both can delay independence, housing decisions, and financial security.
An entry-level role also serves another purpose: it is where experience begins. Consider a hypothetical junior analyst who learns by preparing a first report, having assumptions challenged, and correcting mistakes. If AI handles the initial work, an employer needs another way to develop that person’s judgment.
This is a design problem for organizations. We cannot assume that removing the easiest tasks will leave a functioning path toward expertise. A company may need to invest more deliberately in supervised practice, apprenticeships, and mentoring.
Productivity Gains Don’t Distribute Themselves
There is a compelling positive possibility. A small business could use AI to handle paperwork that previously consumed the owner’s evenings. A teacher could spend less time preparing routine materials. A clinician could have better tools for organizing information, while retaining responsibility for care. These are possibilities, not outcomes guaranteed by adopting a tool.
The human benefit depends on what happens to the time saved. Does it become better service, a manageable workload, or room to learn? Does it become higher output expectations with the same exhaustion? Do customers see lower prices, or do the gains stay within the business?
MIT’s work on early organizational experiments helps explain why implementation matters. Researchers examined more than 20 companies, drawing on interviews conducted between 2023 and 2025. Their recommendations emphasize evidence before scaling, learning when to trust AI, preserving learning and teamwork, and keeping people accountable for outputs. [3]
This is a stronger starting point than assuming that buying AI automatically makes work better.
Beyond Employment: Income, Dignity, and Control
Even a world with strong employment could leave people feeling less secure if pay weakens, decisions become harder to challenge, or work offers less autonomy. Conversely, greater access to useful expertise could expand what individuals can accomplish.
We should therefore assess AI through more than headcount. I would ask whether people have more control over their work, whether new entrants can develop skills, and whether users can understand and question decisions that affect them. We should examine benefits for households and communities alongside productivity within companies.
These are the standards I want this publication to explore. They are editorial questions, not conclusions established by a single study.
Who Pays for the AI Transition?
A research review by MIT’s Guy Ben-Ishai and Neil Thompson argues for a policy response tailored to uneven changes in work. Recommendations include employer-linked training, apprenticeships, and wage insurance. This is a synthesis and policy proposal, rather than a new estimate of jobs lost. [4]
My view is that responsibility should be shared. Employers can measure job quality and invest in credible routes into work. Educators can teach practical judgment alongside tool use. Policymakers can test support for people whose livelihoods change. AI companies can make their economic commitments concrete enough to scrutinize.
Public benefit funds and shared-ownership proposals belong in this discussion. Their value should be assessed against the needs they address and the benefits they deliver. They cannot, by themselves, answer every question about wages, dignity, or participation.
That is why The Shared Upside starts with people. We will follow what AI enables, what it disrupts, and the choices that determine who benefits. The question I want to keep asking is simple: As these systems become more capable, are people gaining a better life?
Sources
[1] https://budgetlab.yale.edu/research/tracking-impact-ai-labor-market
[2] https://digitaleconomy.stanford.edu/news/canariesaug26/
[4] https://ide.mit.edu/insights/ais-impact-on-jobs-demands-a-new-approach-and-new-public-policies/