Skip to content

Supervision does not scale

Ask a programme director at a business school what limits how many capstones, master’s theses or in-company projects a programme can run, and the answer is never demand and never material. It is the hours of the faculty and coaches who supervise that work, one participant or one team at a time, from the MBA to a custom programme for a corporate client. Those hours are rationed in ways schools publish openly, and the rationing is the clearest description there is of the problem.

What schools publish

Rotterdam School of Management’s thesis manual is unusually explicit. Supervision is “usually about six meetings” per student. The committee assesses the proposal and the written thesis within two weeks at the latest. A second assessor reads the thesis but the student will usually not meet them before the oral exam. And the manual tells students that they are expected to take the initiative to ask for and receive feedback from their coach. Read as a capacity rule, that last line means the coach does not chase.

Two former thesis coordinators at Maastricht’s business school describe master’s supervision as complex, challenging and often time consuming, months of one-on-one work, in a workshop for supervisors. Applied projects ration the same way by team size: MIT Sloan assigns one faculty mentor per four-person team and asks the host company to supply a sponsor and a day-to-day contact, in its sponsor brief. Executive coaching inside an EMBA is capped by session count, six at Cambridge, seven at Wharton, taken up by a little over half of students.

Meanwhile the work participants do between modules is substantial and unseen. IMD tells EMBA participants to expect 12 to 15 hours a week on distance assignments about their own company. Cambridge recommends a minimum of 15 hours of study a week. Kellogg-WHU’s EMBA takes about 55 workdays out of the office over two years, and the rest happens at the kitchen table.

Two things no school publishes: the hours allocated per thesis or project, and how many a supervisor carries in a year. In every faculty survey we found, time is the first-ranked barrier to changing anything about teaching.

The arithmetic

Take a cohort of sixty theses. Six meetings each is 360 meetings. A two-week turnaround per draft means a semester has room for three or four rounds, if every student submits on time, which they do not. Add a second reader who appears only at the end. The system holds only because it rations: by meetings, by session counts, by opt-in, and by pushing the initiative onto the student.

This is not a failure of care. It is what a mailbox forces. Drafts arrive whole, out of order, at deadlines. To find the fifth of the cohort that needs a supervisor this week, the supervisor reads everything. Nobody knows who is stuck until a deadline passes. And generative AI has made the drafts longer and smoother, so each read takes more time and tells you less, a point Ivey’s Tony Frost makes about case preparation in Poets&Quants: students now arrive with positions rather than questions.

Supervising by reads

The way out is not more supervisors. It is to change what a supervisor’s hour is spent on. Three moves do it.

First, the method is set once. The steps of the project, what each step must contain, and where a signature is needed, written down for the whole programme rather than carried in one person’s head.

Second, the first reading is not the supervisor’s. Something that never tires reads every draft against those criteria, tells the participant what is missing, and asks for a revision before a person sees it. Most drafts go back at least once without costing anyone an hour.

Third, the supervisor reads what is ready, in a queue sorted by who has waited longest, and decides. Return with one note, or sign. The number that matters becomes reads per supervisor per week, not meetings per student, and a supervisor can carry a cohort that would have needed three.

The decision stays human throughout. Nothing in this replaces the judgement that makes a signature mean something. It removes the chasing and the first pass that were never judgement in the first place.

What a remedy must have

  • Criteria before drafts. If the bar is not written down, nothing can read against it.
  • A first reading that is not the supervisor’s, and that returns the draft with what is missing rather than a score.
  • A queue, not a mailbox. Who is at which step, who is waiting, for how long.
  • Decisions on record. Every return and every signature, with its note.
  • No skipping. A participant cannot reach step four without step three being signed.

Where Ulern stands

We build Ulern, the applied-learning platform for business schools and executive education, so read this section knowing that.

In Ulern the supervisor sets the steps and the criteria once. The AI tutor reads each draft against them, tells the participant what is missing and asks for a revision. The supervisor sees each participant and each cohort by step, with the ones waiting on a person first, and one email each morning lists what waits on them. They return a step with one note or sign it, and everything is recorded. We have not yet measured hours saved per supervisor on a live programme. The first programmes to run this way will give that number, and we will publish it.


This is the third of seven pieces in Assign and hope, a series on what goes wrong while a programme’s project runs, and what a remedy must have. The others: Did they write it, or did the AI? · Activities are not outcomes · The 90 days after the programme · What corporate clients now ask of schools · The thesis in the age of AI · What to ask an AI tool before December 2027

← All posts