Real Alpha
Watch modeSubscribed

The Tim Ferriss Show · 3 min · Dec 17, 2025

Vertical AI startups can defend workflow niches

Bill Gurley says startups should avoid the frontier and target overlooked industries with local data and embedded workflows. Those integrations are harder for foundation-model companies to replace.

Bill Gurley

Bill Gurley is general partner at Benchmark and author of Runnin’ Down a Dream.

Backed and sat on the boards of Uber, Zillow, OpenTable and Stitch Fix.

  • Tim Ferriss, Host · Host, The Tim Ferriss Show

    Wrote The 4-Hour Workweek; an early investor or advisor in Uber and Shopify.

About this segment

Bill Gurley says startups should avoid the frontier and target overlooked industries with local data and embedded workflows. Those integrations are harder for foundation-model companies to replace.

From “Bill Gurley — The AI Era, 10 Days in China, & Life Lessons from Bob Dylan, Jerry Seinfeld,, and More”, Dec 17, 2025. Watch the full episode

Answers “Is SaaS dead?” See all 9 answers

“there are workflows there are data sets that are local to your customer and that has stuff has to be stitched together”
Bill Gurleyat 15:16 · The Tim Ferriss Show15:16

Takeaways

  1. Stay away from capabilities at the frontier of major AI labs.
  2. Look for industries with customer-specific data and workflows.
  3. Integrated software can defend more than a model answering questions.

In this episode

Full episode
  1. 5:47AI deal structures and SPVs raise risks for investors5:21
  2. 53:57Being near the action improves learning and opportunity5:04
  3. 1:31:37Use regret and curiosity to decide when to change paths5:34

  4. 13:43Vertical AI startups can defend workflow niches3:20

  5. 18:35China's provincial competition can produce rapid innovation3:16Up next
  6. 1:07:37A small extra contribution can create career access3:21
  7. 1:45:27Publicly funded research could require open-source rights7:01

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

  1. 13:56

    how would you think about maybe you can give a hypothetical example of looking for someone who has very very sophisticated domain expertise and experience who's now intersecting with ai and has a unique because of the combination perspective on things to invest in as an angel investor separate that from something that's just going to be consumed by the fundamental models and these larger companies that from a career perspective or from an angel investment perspective how would you pick folks you don't think are just going to end up working on something that gets replicated in short order by the bigger companies the key is just to stay pretty far away from the edge of whatever you can go online and see interviews with the people in anthropic or OpenAI and what they're working on. Like if it's the next thing they're going to do, I don't think

  2. 14:50

    you're going to be protected. But as I think about founders and angel investors, you're talking about a pretty broad array of things. At this point, as I mentioned earlier, you're not going to back the next big model company. Besides, if you were, you need a billion dollar angel investment to go make that happen. The game's changed. There's so much money involved. I think you're going to want to be off the beaten path anyway when i think about these deeper verticals i don't think it will make sense for open ai to go crush every little vertical waste management even if the model's capable of understanding that subject matter there are workflows there are data sets that are local to your customer and that has stuff has to be stitched together so i think having an understanding of a particular industry and one that's not going to be on the next thing to do

  3. 15:46

    list at OpenAI would probably be your best bet. Got it. So is it fair to say, if I'm understanding you correctly, that effectively looking for something that would not be a high priority for one of these larger companies and also proprietary data set of some type? Proprietary data sets, the more kind of workflows that exist. These are better because you can build software around those things. What is a workflow?

  4. 16:09

    The thing that popped in my head, I'm on the board of Zillow. Zillow's been investing for the past five years in tools that help the realtor do their day-to-day job. They have a tool called Showing Time that helps you book in-person tours at houses, as an example. There's putting the mortgage together, getting the sign-offs on. There's just all these tasks that have to be happening. can be automated and so tasks that can be automated that can be integrated with ai the more of that stuff you can build into a system the better off you're going to be protecting yourself from a model that just answers questions which is why i brought it up let's move on to big topic