Conversations about the MBA tend to circle tuition, rankings and post-graduation salary. Two questions matter more and get asked less: what actually happens inside the classroom, and who actually gets admitted.
Two pieces of reporting published in 2026 answer both with unusual specificity. One is an account of a Stanford Graduate School of Business class where students do not discuss AI — they build AI agents and set them negotiating against each other. The other is a profile of 117 soon-to-graduate students nominated by their own schools, which shows that the person getting into a top programme looks rather different from the stereotype.
A class that makes students build AI rather than debate it
The course is taught by Andy Hall, the Davies Family Professor of Political Economy at Stanford GSB and a senior fellow at the Hoover Institution. Roughly 30 students, with a live experiment run in class every week.
The assignment is not an essay on AI governance. It is this:
- Train an AI agent to represent your own views. Students go through a preference-elicitation interview, and the agent is built from what that process extracts.
- Test whether the agent actually understands you, by checking its choices against that student's real voting record.
- Release the agents into a simulated legislature, where they negotiate, deliberate and trade votes on shareholder proposals. Results appear on screen in the room as the experiment runs.
The tooling is ordinary and publicly available: the Claude Haiku 4.5 model accessed through OpenRouter, plus Claude Code. No proprietary university infrastructure.
What the class actually teaches shows up in what the students watch happen. Behaviour nobody programmed emerges on its own: strategic voting, vote trading, multi-party bargaining. So do the failures — moments where an agent clearly misrepresents the person who trained it. Hall's argument, in short, is that you learn far more about agentic governance by building it than by theorising about it.
For anyone weighing an MBA, this is a clear signal: leading programmes are shifting from teaching about technology to building it during class hours. Hands-on fluency with AI tools is becoming an expectation rather than a differentiator.
What the graduating class of 2026 looks like
Poets&Quants published its MBAs to Watch list for the Class of 2026 on 6 September 2026. The method: contact 83 full-time MBA programmes, receive 217 nominations, and select 117 students from 70 business schools.
| Measure | Class of 2026 |
|---|---|
| Students selected | 117 |
| Schools represented | 70 |
| Gender | 60 men · 57 women |
| Nationality | 66 American · 51 international |
| Programmes contacted | 83 |
| Nominations received | 217 |
Almost 44% of the group is international. That is worth noting if you are applying from outside the US: the cohort schools chose to put forward is not heavily weighted toward domestic students.
Who hires them
Among these 117, the four largest recruiters are all in consulting:
- Boston Consulting Group: 9
- Bain & Company: 5
- McKinsey & Company: 4
- Deloitte: 4
The three big strategy firms alone account for 18 of the 117. If consulting is your reason for doing an MBA, that route is demonstrably still wide open.
What they have in common
The recurring characteristics are not spotless transcripts or tidy linear résumés:
- Non-linear pre-MBA careers — industry switches, role changes, detours
- A substantial contingent of military veterans
- Pivots into consulting, investment banking and technology after the degree
- Entrepreneurial ventures or hands-on product building
- Community and social-impact work
Put differently: admissions committees are not looking for a flawless record, they are looking for a story with a direction. Three years in engineering, then operations, then a move toward consulting — if you can explain why — reads stronger than a polished profile that says nothing in particular.
What both pieces suggest about your application
- Do not hide the turns in your career. Most of the nominated group took a crooked path. The issue is never that you changed direction; it is whether you can articulate the logic of the change.
- Hands-on AI experience is now shared vocabulary. When a Stanford course asks students to build their own agents, the specific things you have done with these tools become genuinely usable material for essays and interviews.
- Consulting remains the widest exit. If that is the target, weigh schools by how present the consulting firms are in their employment reports, not by overall rank alone.
- But the entrance is still the test. None of the above matters until your file clears the first filter, where a GMAT or GRE score is the first number the committee sees.
If you are still in the preparation phase, the first useful step is knowing where you actually stand — from a scored practice test, not from a feeling.
Sources
- Reporting on the Stanford GSB class building AI agents — Poets&Quants, 30 April 2026
- MBAs to Watch: Class of 2026 — Poets&Quants, 6 September 2026
The application takeaways are Clever Academy analysis.