(2024-01-20) Me Contra ChatGPT for Product Management

Can ChatGPT help with product management?

Person1 posted: we size customer pain points and feature ideas by using AI to extract and aggregate this information from recorded sales calls, success calls, CRM notes, support conversations and surveys

Person2 posted: this seems like a good idea, and lots of people are pitching it, how come none have become Big?

My answer of the moment, which I thought I'd share:

Because they’re bullshit, they’ll give you “faster horse” or “higher quality for lower price” recommendations. They make it easier/faster to generate plausible answers/solutions, but the odds of those being successful solutions doesn’t improve.

In the same way that “AI as writing assistant” makes it easier to generate an essay which the reader will immediately forget.

It’s like solving the problem through group discussion. Possible outcomes are

• the discussion is badly handled and you get nothing useful
• the process seems pretty good, you generate lots of ideas, but none of them really Solve the Problem
• the process seems pretty good, you generate some ideas, and 1 of them really Solves the Problem.

AI lowers the cost of achieving Outcome2.

“Solving” product-market-fit is an unbelievably hard problem. Some say every successful company was mainly Lucky, one of their random guesses actually worked.

Maybe there’s a process where AI helps you generate many ideas, and the Right Person can notice the 1 Real Solution (or get inspired to generate the 1 Real Solution which the AI didn’t mention).

But if you don’t have the Right Person driving, then you’re back at Outcome2. Which is solving the Wrong Problem. (cf There are 3 Games of Product Management)

Remember, all ChatGPT/LLMs do is generate reasonable-sounding sentences by gluing together phrases recorded by other humans. There is no actual intelligence there, no fit-to-context, etc.

(i could always be wrong)