Why I'm building an AI assistant

There's been a lot of buzz about artificial intelligence this year - it would be difficult to have missed it, seemingly penetrating into every sector and walk of life. And yet, the most common question I hear from people outside of tech asking about it is 'but how does it help me?" or "but what can it do that's actually useful".

Most people outside of the tech bubble still see it as largely a toy (which incidentally isn't a bad thing), or even if they do not, have not found ways to reliably integrate it into their work and life.

So the question is; what is the biggest barrier to artificial intelligence being adopted in a meaningful way by every individual and every company. I believe the biggest barrier is a product one, not a technological one.

What do I mean by that? I believe that the current generation of large language models (LLMs) already hold significant power, the majority of which has not been utilised in real-world applications. And so the valuable area to focus one's time is at the product layer - understanding use cases, and building an architecture on top of LLMs to solve problems within these use cases as best as possible.

It is for this reason that I think Seth Rosenberg's (Greylock) 'Product-led AI' thesis is the most aligned with my own views, and describes the nature of the largest companies we're going to see emerge in the space in the coming years. It's also why I believe incumbents are not as well placed in productising AI (i.e. delivering it direct to consumers, as opposed to building tools for the builders) as many say (because then they'd be counter-positioned against themselves in many cases).

So why am I building an AI assistant?

The answer falls into two reasons:

  1. In building my personal holding company, I found there were a lot of personal use cases that I had for integrating artificial intelligence into my workflow, that didn't yet exist
  2. Over the past couple of years, I've realised that 'product engineering' describes where I feel I'm most effective best

Starting with the first point: the products that I've found in the market lack the quality that I would expect. To an extent that's to be expected because the sector is new and developing, but it seems to have taken longer than I expected, and so I started building something myself for my own use case. Speaking with others about this, they've mentioned they'd like access to the product as well, and so began my process of converting the hacked-together solution I'd made for myself into a more mature, production-ready application.

On the second point: I taught myself to code at a young age, and have tried to remain on top of the emerging approaches for developing applications. But I wouldn't claim to be the world's greatest programmer, nor would I want to be. Where I feel most effective (and enjoy the work most) is writing code to move towards the best product solution, releasing it, and iterating on that process hundreds of times. This feels like the perfect context to do just that; the technology itself now exists (though it will also improve and become cheaper over the next few years), with the majority of the work now being centred around building the best user experience, and designing solutions for specific problems and use cases.

What I've built

So far the product helps me to manage my email and calendar, and I've started to intelligently automate some of my most repetitive workflows. When the product goes live, it will extend beyond this to act as someone's personal assistant in a wide number of use cases.

Subscribe to Charles Williamson

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.