This past year I’ve had many technology startup founders, along with others that have an idea for an offering, ask me what I think about their idea of using ChatGPT to power their idea to provide their users with generative content to use in some way. A working example of providing generative content might be Canva and their integration. These startup ideas have tended to fall into three buckets based on their complexity:
- Simple Prompt Generation. Essentially creating an interface with pre-ordained prompts so users don’t have to concern themselves with prompts. My feedback with this is that you are 100% dependent on a 3rd party technology provider that you don’t have any control over and because of the risks associated with that, you’ll likely not be able to fundraise, and any exit multiples will be low.
- Prompt Generation as Large Part of, But Not All of, an Offering. Similar to #1 above, these tend to be fairly simple, but often take the generative output and mash it with something to provide some value add. Because these business ideas still die if ChatGPT cuts you off, I give the same advice as #1.
- Generative Content Integrated Into Larger & More Complex Offering. These are ideas where the generative content is integrated throughout an existing tech stack and/or organization. Think ChatGPT helping a payroll manager do their job more efficiently. Perhaps Canva fits into this bucket – they had a thriving business long before ChatGPT and if their integration went away, they’d still be in business. My feedback in these cases is that because you aren’t 100% reliant on a 3rd party and can meet your customers’ needs without that 3rd party, then you should be able to fundraise and eventually realize strong multiples.
Now my two questions for this group are:
- Do you agree with the above categorization and the implied funding and exit outcomes, or do you view it differently? Have you seen anyone in #1 above successfully raise VC funds?
- With ChatGPT now doing voice, and GPTs (where you train your own GPT on your own data to enable domain-specific expert output), do these new services change the recommendation calculus in the above 3 scenarios?
Seems to me that voice does not change the calculus, it is just a different input and output format (although super cool and useful).
But creating your own GPTs that leverages your own proprietary data/content that no other organization has access to, as part of a more complex system, I think falls into #3 above, including the finance opportunities.
As an enabling technology, generative AI is going to create many new opportunities for technology entrepreneurs and I want to make sure I understand when and where it creates fundable, high growth companies, and when it is really just a good feature.
This group is terrific, and I really appreciate your feedback. Thanks in advance!
I think for #1 and #2 above you are missing the value of open source and alternatives. OpenAI is currently really good, but…
You have LLama 2, Falcon, Mistral, and others that are open source and that you can go download today, for free and use commercially. So, your prompt engineering will likely work for them as well with only minor adaptations.
You will also have closed versions that will be much more productized. MSFT, GOOG, Amazon, Anthropic, etc are all working on theirs and they will certainly have API calls as well.
So, there will be less OpenAI dependence in the future.
Yes, for sure, as other alternatives develop this becomes less of an issue.