New software founders often ask me to review their funding pitches or their startup strategies.

I let them talk and I ask questions. Then I tell them “This is what it sounds like to me. Did I get that right?”

They usually talk for 15 minutes and I summarize, in my own way, in 1 minute.

I don’t repeat back what they said as they said it.

I repeat back what I heard and what would be more useful–based on hearing many thousands of startup pitches and pitching successfully myself. After 30 years of creating real-life marketing and sales messages, getting results, and studying the underlying structures of communication and persuasion.

Their usual response is “Oh, that’s good. Can you repeat that so I can write it down?”

If it takes a few tries to get it right, it usually means something is missing or confusing about their own description. More questions.

I call it GregGPT with my AI–Actual Intelligence. That gets a knowing laugh.

Will a ChatGPT of the future be able to do this same thing?

There’s a high likelihood that it will. I don’t know when.

If a specialized large language model has ALL the data, all the pitches and what worked, all the structures of language, and all the real-time reactions, then it should be very good at this.

Artificial Intelligence with years of learning and complete data should be better than Actual Intelligence with our biases and imperfect data.

I’d want my doctor of the future to be consulting an AI assistant that uses a million times more data than my doctor can process, without the normal human biases and shortcuts.

There will always be a human element to this. Learning to prompt in the right way, modifying the result, and presenting with skill. Humans have to hear it to do something with it, too.

Until my AI talks to your AI. Maybe that will happen.

This is a new trajectory for powerful technology that will create a lot of change in the world.

Where do you think this is going with AI for communication challenges like this?