The AmplifierReal cases · Student · MBA
Tanvi, 23: a case competition, four teammates, and a deck built by a tool at 2 a.m.
Forty slides nobody can defend, a judge who asks one question, and a team that has never agreed what it is recommending.
Planned for release phase 2. The page is complete as a draft; interactive parts run in the browser and do not yet save to an account.
In brief
The situation. Tanvi is in her first year of an MBA in Pune. Her team of four has a national case competition in nine days: a struggling regional airline, a 60-page case, a twenty-minute pitch. Last time, a teammate fed the case to a tool and it produced forty polished slides that recommended everything. A judge asked, "What would you cut first?" and four people looked at each other. Tanvi has been asked to lead this time.
Step 1 — Define the End State
"One recommendation, defended for twenty minutes by any of the four of us, that a judge who has run an airline finds specific to this airline. Fifteen slides. When they ask what we'd cut first, we answer in one sentence." Audiences: one judge who has run something, and the teammate least confident on her feet.
Step 2 — Map the Flow
Her list: read the case (all four, alone), each write their one-paragraph recommendation before any discussion, argue to one, then analysis, then slides. AI review adds: a list of the questions an airline operator would ask that a student would not; a check of the case's numbers, because cases contain traps. It suggests generating the deck first "to have something to react to". She declines: that is how last time happened.
Step 3 — Assign Tasks
The recommendation: humans, argued out on a whiteboard. Generating the operator's questions: AI, then each teammate takes three to answer. Checking the arithmetic in the case: AI, verifiable. Building the financial model: hybrid — the tool drafts the structure, a teammate builds it and can explain every cell. Slides: drafted from their outline by the tool, then every slide rewritten by the person who will present it.
Step 4 — Select the Tool
A conversational tool with long context for the case document; a spreadsheet for the model — not a chat tool. The presentation stays in whatever they present in.
Step 5 — Prepare and Direct
Criteria: every slide has one claim and one number; any slide the presenter cannot explain without notes is cut; the recommendation fits in one sentence on slide two; the "what would you cut first" answer is written down before the deck exists.
Steps 6 and 7 — Execute and Evaluate
The operator's questions are humbling — eight of twelve they cannot answer from the case, and three turn out to matter. The tool's first slides fail the one-claim criterion almost uniformly; the rewrite by presenters is where the deck becomes theirs. The number check finds a load-factor figure in the case that does not reconcile — a trap, and they use it.
Step 8 — Integrate and Stress-Test
Dry run in front of a professor. Small fixes: two slides in the wrong order, one chart unreadable. Structural: the model's assumptions and the slides' assumptions differ on fuel cost — a Step 3 problem, two teammates working from different pages of the case. Reconciled, model re-run. The "cut first" answer stays one sentence.
What changed
Fifteen slides. The judge asks what they would cut first. Tanvi's quietest teammate answers in one sentence, because it was written down on day two. They do not win. They come second to a team the judges praise for the same thing.
Everyone writes their recommendation alone before anyone speaks. Generate the hard questions before you generate anything else. Whoever presents a slide rewrites it.
A composite case, not a client engagement; the person is invented, the situation is not.