A competition between Wharton MBA students and ChatGPT that judges generating ideas for a new product or service finds that 35 of the top 40 came from ChatGPT
We put humans and AI to the test. The results weren't even close. — How good is AI in generating new ideas?
Context & Ripple Effects
This is a sharper follow-on to Wharton’s earlier finding that ChatGPT could pass an MBA final exam at a B-range level. It shifts the comparison from answering course questions to generating commercially oriented concepts.
The result also lands amid broad workplace experimentation with the tool, while a separate study found it improved both the speed and assessed quality of certain writing tasks in a controlled writing-task study.
First-order effects
- ChatGPT’s ideas occupied 35 of the competition’s top 40 slots, giving Wharton students and instructors a concrete benchmark for where unaided ideation now faces automated competition.
- For participants, generating a first set of product or service concepts becomes less of a differentiator; evaluating, refining, and selecting ideas becomes more central.
Second-order effects
- Employers already testing ChatGPT across roles will have added reason to incorporate it into early-stage brainstorming, rather than limiting use to drafting and summarization.
- Business-school and entrepreneurship workflows may put greater weight on prompt design, market judgment, and validation, because a high volume of candidate concepts can be produced quickly.
Third-order effects
- If similar results recur across idea-generation tasks, generative AI could shift knowledge-work advantage from producing an initial answer to owning the workflow that turns many candidate answers into decisions and execution.
- The durable question is less whether models can generate plausible ideas than whether institutions can reliably distinguish novel, viable concepts from fluent but weak ones; this competition alone does not settle that.
The trend: Generative AI is moving from a drafting aid to a workflow-native partner for the early, high-volume stages of knowledge work.