“Conviction” is hubris

Thanks to AI, it’s never been cheaper and easier¹ to build the wrong thing.

But I think some founders and executives have used this moment as permission to do what they’ve always wanted to do, which is to skip all this “discovery” and “testing” nonsense, and instead direct teams to implement their idea, based on their “conviction”.

One day in the future, they might even be right, if/when the cost of turning an idea into production quality software is measured in days.

But in today’s reality, it’s not, for an organization of any real size and scale (say, more than 15 employees and thousands of users).

Designs still need to be thoughtful and polished; tricky interactions are still tricky to get right; code architecture, performance, and security are still critical and non-trivial; bugs still need to be fixed; product analytics still need to be defined and dashboards created.

And, when it’s all done and launched, you still need to wait weeks or months to find out if the feature was actually worth it.

At the end of the day, maybe we’re launching new features in 4-6 weeks instead of 8-12 weeks. Tack on 2-4 weeks post-launch to find out if the new feature worked. So even with a 50% increase in velocity, we’re still looking at 6-10 weeks. That’s an expensive bet to make based on “conviction”.²  

Even in our new AI world, the most expensive mistake companies can make remains shipping the wrong thing. Take a week to do your user research, talk with customers, user test prototypes, and conduct painted door tests, and iterate based on what you learn.

We’re so lucky to live in a world where we can use AI to learn even faster. So let’s actually do that, and not regress into the world of 20 years ago where the only tool we had was “launch it and see what happens”.

The Workshop

This is a newsletter-only section where I share a half-baked idea in hopes that y’all who are smarter than me can work it out with me.

As a person who has always loved the class of user problems called Discovery (browse, search, recommendations, UGC), it is such an exciting time.

Before LLMs, I think we’d been kinda stuck in a rut, squeezing 2% gains out of conversion funnels we’d been optimizing for years.

But now, there’s so much we can do!

We can finally help people think through what they are looking for when it’s still early in their process and things are unclear or unknown.

We can do such a better job at search, when the keywords they are looking for aren’t exactly matching what’s the catalog text.

We can be multi-modal, taking voice / photo / video as input and also translating output into voice / photo / video.

There’s so many new UX patterns to be discovered and tested, I’m excited for how much better we’ll be able to help people. And what a great time for companies who are in that step of the user journey to exert strategic (and revenue generating) influence.

For any of my friends who are in this every day, I would love to hear from you: what new patterns are working for you?

1  Turning an idea into a working prototype is super fast. Agents can generate thousands of lines of code, and invoke other agents to code review it and test it.

These are real velocity increases, maybe 2-3x faster to get to a throwaway prototype, maybe 30% faster overall to get something production quality shipped.

2  A sign of our bubble times is how many startups, fueled by founder “conviction” and too much VC funding, will run themselves right off a cliff.