Agreed! I should clarify that learning the job doesn't mean you can actually do the job. You might feel like you can but you probably can't. There's a reason why most value from PE comes from simply cutting expenses.

O and developing relationships in that industry, even validating something is a real problem is hard without knowing somebody who can go out of the way to help you understand real workflows.

Thankfully ~every vertical AI founder I've met is very immersed in their industry. They know learning the job comes first.

13y of experience watching my wife's dental offices, thinking about lending a hand on 3 different occasions, and I can confirm there are 724 layers to the onion that you start to peel back. That said, we're absolutely going to do this at some point the next few months/years.

Yes exactly, I’ve spent three years learning about running a life insurer than complexsystemspodcast.com/ep…

💯 agree but also 6 years later mortgage servicing is very well covered by Valon. Highly complex industries are a huge moat — most people won’t take the time to really learn them (and not doing so is a sure path to failure).

Yep. the domain complexity here is the real moat/barrier. it's not just the data, it's encoding all the edge cases, exceptions, and regulatory spaghetti that ppl have built manual processes around for decades. AI models struggle w/ this structural, procedural knowledge way more than ppl realize

they will take the beatings that private equity people took 10 years ago and will pay for it with a more wonky capital structure

we do this for alternative asset managers mainly real estate and PE, and occasionally get a hedge fund client having us do an AI tool and frankly, we just do a worse job because we cannot empathize as much. this I say as someone who worked in asset mgmt space all my life

Most skip the hard part: deep, boring expertise. That’s why they lose to the quiet operator who already knows the rules.

Most people think "automate the whole thing" when in reality it's about increasing throughout and quality of the process. You go from x cases/person/day to 10x cases/person/day with better quality.

One of the founders should be in the industry.

And also feel comfortable hiring some mortgage services officers! The customer doesn’t care if it’s AI or human it just has to work. The X% that isn’t automated or accurate enough with AI still needs to work or the TAM shrinks dramatically.

This talk about learning workflows reminds me somewhat of “expert systems” books from the 80s, all the ones I’ve seen talked a lot about that

I am shocked that this is exist. How could they don’t understand and learn how things work and interact in details? Then what are they fixing/solving? If thats true, these type got successful by pure luck.

Rip mortgage servicing startups. Get sued

This is PE job, not VC.