A $699 purchase of access to an AI faculty clone does not prove customers value the clone. It may prove they value the institution, the faculty name, and the signal that comes with buying from Harvard.
In 1985, Coca-Cola chief executive Roberto Goizueta faced a result that looked unusually clear. Blind taste tests showed consumers preferred a new, sweeter formula to the original Coke and Pepsi. The company had evidence, a competitive threat, and a product that appeared to win when people judged the liquid alone.
Then Coca-Cola replaced its original formula.
The backlash exposed a variable the taste tests had removed. People were choosing more than flavour. They were choosing memory, identity, habit, and the meaning attached to the Coca-Cola name. The company brought the original formula back as Coca-Cola Classic 79 days later.
Mark Pendergrast documents the episode in For God, Country and Coca-Cola. The important part for a founder is not that Coca-Cola made a research mistake. The tests answered the question they were designed to answer. Customers preferred one taste in a blind comparison. Coca-Cola treated that answer as evidence for a larger decision.
That leap is where the $699 trap begins.
The institution changes what the purchase means
Put the same AI faculty clone on an unknown founder’s landing page and the buying decision changes.
A customer considering Harvard’s offer may expect access to a respected professor’s thinking, the confidence of a known institution, a credible line on a professional development budget, or simply the chance to inspect how Harvard is approaching AI education. The underlying mechanism may still be useful. The purchase alone cannot tell you which part carried the decision.
For a founder in Accra, Lagos, Berlin, or New York, this distinction matters because the institution wrapper is difficult to reproduce. You can build an avatar that answers from a body of lectures. You cannot copy decades of trust, alumni identity, faculty reputation, and employer recognition into the product.
The Coca-Cola tests isolated taste and lost context. A founder copying this offer risks doing the reverse: observing a purchase surrounded by institutional context, then attributing the result to the technology inside it.
Separate mechanism demand from authority demand
I would test the mechanism before committing runway to the full product.
Start with the job the buyer expects the clone to perform. Perhaps they want a fast answer from a defined body of expertise. Perhaps they want help applying a professor’s framework to a live company problem. Perhaps they want the feeling of private access to someone whose time they could never book.
Those are different products.
A useful test removes as much borrowed authority as possible. Offer a small group access to an AI adviser trained on material you have the right to use. Give it a narrow promise, such as reviewing a pricing decision against a published strategy framework. Charge for the result. Then watch what happens after the novelty wears off.
Do users return with a second decision? Do they cite an answer in a team discussion? Do they ask for a human review because the AI response feels too general? Do they pay when the expert’s name carries less status?
A waitlist cannot answer those questions. Neither can a free demo filled with curious founders. A paid pilot with a defined decision and a follow-up interview gets closer. That is the same pressure behind how Daniel tested a $699 strategy before spending runway.
Conflicting answers reveal the actual product
The interface will attract attention first. Reliability will decide whether the product survives.
If two customers ask similar questions and receive conflicting guidance, the founder has to explain what the clone represents. Is it retrieving the faculty member’s documented position? Synthesising source material? Generating a plausible answer in a familiar voice? Those differences affect trust, evaluation, and liability.
The institution may absorb some early uncertainty because buyers already trust the name. A young company has less room. It needs visible source boundaries, a clear account of what the model knows, and a path for handling answers that exceed the material.
That problem deserves testing alongside demand, not after launch. I explored the operational side in what happens when an AI faculty clone gives two students conflicting answers.
Make the next test harder to misread
The next experiment should force a choice that separates curiosity from value.
Pick one decision your target customer already faces. Build the smallest version that can help with that decision. Recruit buyers because they have the problem, not because they recognise the expert. Ask them to pay enough that the purchase competes with another real use of their budget.
Then measure what they do after receiving the answer. A repeat use, a referral tied to the outcome, or a request to bring the product into a team workflow says more than admiration for the demo.
Coca-Cola learned that a preference measured outside its full context could not support the decision management made from it. The AI faculty-clone market presents the same risk from the other direction. The context may be doing more work than the mechanism.
Before building the clone, remove the crest from the experiment. See whether the advice still earns the $699.
Comments
No comments yet.