We should not limit ourselves to being AI users; we should help develop the technology and models that best position us to solve the problem we are addressing.
If I had to describe what it’s like to start a startup with AI, I would say it has two parts: applying AI and developing it. From our perspective, the way to use AI has always been to think about what problems we can solve more scalable using LLMs or Gen AI. The flip side of this question is a historical one for all startups: how do I increase my results with the same resources?
But there’s another angle I’d like to bring up: we need to start from the premise that any problem that can be solved with AI will be solved with AI and ask ourselves where this leaves our startup. Another way to phrase this question is: if I started solving this problem today with the available AI tools, how would I do it?
I found it very inspiring to see this week at the annual Figma conference how they are thinking about AI. To give some context, Figma is a 7-year-old startup. Their first innovation was to show that it was possible to design interfaces (UIs) in the browser in the same way it could be done with Photoshop. Today, 7 years later, they are asking this question, and their answer is a product that designs interfaces through a prompt. (In the process, by announcing this, they are disrupting the innovations of many new startups).
Figma is asking itself this same question and is developing a product as if it were seeing the problem of creating interfaces for the first time. And one more thing: they are leveraging their current infrastructure and user base to position themselves as the winner in the AI-driven interface design market. Those of us who are not Figma have to ask ourselves: where does this place us? What can I learn from this, and what should I be doing that I am not?
Going back to the beginning: I think that to create a startup with AI, it is important to apply what exists and help develop it. (This is not new. In technology, we have always worked with tools while helping to develop them: I refer to open source and the great tools we all use like Node or Nginx). I believe we should not limit ourselves to being AI users; we should help develop the technology and models that best position us to solve the problem we are addressing. It's that, or trust that someone else will do it.
