Startups: To be or not to be dependent on AI APIs

Extended from original post on LinkedIn as a quoted repost of Alex Issakova’s post on the AI winter.

The next AI Winter has been one of my main alerts.
It sounds very compelling to chase the hype and ride the wave without looking back.
But that’s not reality.
Currently too many companies are promising “all in one” AI solutions that, either or both

1. don’t really generalize to “all”, as usual;
2. are good at “one” but too expensive to operate independently and there’s no realistic horizon for them to actually become cost-effective.

I really started to realize how critical and significant this was after taking the role of directing a new technology stack from day zero, making all core decisions and driving our tech vision.
I realised that for every decision we’d be making, we could either be realistic about the direction and the time it would take to get us there, or instead, to jump on the boat of over-promising by over-riding third-party solutions which aren’t guaranteed to succeed.

You get to pick:

• either build a solid foundation during 1-2yrs for your company that doesn’t seem to attract that many investors or catch the attention of that many clients;
• or go full showbiz mode in 2-3 months and pump up your marketing over actually spending your time, focus and resources in your technology.

In the later case, you get to show off a neat party in your nice castle except it’s a house of cards that may topple at any moment.

Strangely, it seems that investors, who should be quite more risk-aware, have actually been falling for all those party castles.
After all this hype is over and investors start realising how much money they’ve lost, we’ll be looking around and the only success stories will be the ones where the focus was economic and technical sustainability.

Winter IS coming, so the question is, how do you prepare for it?

• Adapt your product lines towards intermediate monetisation steps instead of going solely for the shiny big new thing.
• Ensure you have a clear bill on your operation and that you HIRE people who can help you understand if you really need a bill that size or if there’s anything that can be done (including running some services yourself).
• Ensure you focus on User Experience - that will be the life jacket of the next AI winter.
• Focus on solving specific problems and delivering specific added value, while aligning that all with UX - you’ll only survive if you ensure that customers will actually want to pay the real price of your operation.
• Oh, UX IS NOT GRAPHIC DESIGN. It’s about aligning your product with the mindset and workflow of the users. UI is a significant part of it, but once your product is an interactive experience, UX quickly shifts strongly from the UI design aspects to the Interaction Design aspects.

Last but not least:

• Ensure you prepare your own fallbacks. How much are you currently paying for AI sevices? AI coding assistants? While it’s nearly free - use them to learn how to build your own systems so that you can have fallbacks available for when your current services either shut down or increase their cost 10-100x.