Find an un-sexy domain that you have more access to than the average person. Start to build domain expertise in that area as quickly as you can (people are surp

Find an un-sexy domain that you have more access to than the average person. Start to build domain expertise in that area as quickly as you can (people are surprisingly willing to talk when you don't want to sell them something, but just learn about how they do things). Build something (anything!) that uses some cool ML technology (or even just some simple statistics) in that domain. Think of this as an ultra-MVP - something you build part-time in the order of weeks, rather than full time over months. This keeps your "investment" low and makes you mentally ready to accept changes. Loop back with the people in the un-sexy industry to get feedback. Remember, not all industries are bombarded with technology - you'll need to strike a balance between showing them something sufficiently "fancy" to pique interest, and abstracting away your technology so they focus on a problem it solves. When my cofounder and I started ~4 years ago, we chose an un-sexy industry that my cofounder knew from years of consulting. This was awesome, but even with his amazing experience, our initial idea for a product didn't take off (which was fine - it doesn't take too many wins to make the weeks of development/sales profitable). Over the next 4 years, we built another couple of offerings, which sold better than the first. They took hold in various ways, but the insights behind each product were usually driven (or at least influenced) by the people we talked to about our initial product. Which brings me back to the key - build things in a low-cost way and use that to identify tangentially related problems until you think you've found a big enough pain point.