---
title: What to call me about
template: numbered
order: 3
kicker: Problems
description: Seven problems Ben Siverly has worked on, from forms people don't finish to money that has to be right, where he met each, and one fact from his résumé.
deck: Seven problems I've worked on, where I met each one, and one fact for each, quoted from my résumé word for word.
teaser: Seven problems I've worked on, where I met each one, and one fact for each.
updated: 2026-10-29
art: street 128 7 260 260
close: If one of these sounds like your week, write to me. The first conversation is free.
llms: seven problems he has worked on (people who start and don't finish, payments that have to be right, measuring what a change did, AI plans, trying new tools, customers nobody has talked to, a first product person), each with where he met it and one fact, quoted from his résumé.
---

## People start, and don't finish

People open the form, get partway, and leave. They can't save it or come back to it, and nobody can see where they stopped.

At AppFolio I rebuilt the rental application end to end so it worked like a modern app: accounts, so applicants could finish later, and reporting, so property managers could see what was happening. Submissions went from **44% to 70%**.[^1]

## Money has to move, and it has to be right

A payment fails halfway through checkout. The order stalls, and nobody is sure whose move it is.

At AppFolio I owned checkout for application fees. I designed an assumed-success window and a recovery path for failed payments, and application-payment completion reached **99.1%**.[^2] Before that, at Lively, I built the money-movement platform: transfers, processor integrations, ledgers.

## Nobody can say what a change did to the numbers

A change ships, and a month later nobody can say what it did. The report that would tell you gets rebuilt by hand, when someone has time.

At Lively nobody could see how product decisions moved the core metrics, and there was no data-science function. I stood up the first one, and the data-services platform under it, and the work became the roadmap. At Oracle I built a system to run hundreds of evaluations against model updates, so a change was judged on real statistical significance, not gut feelings. I also automated a data pipeline that had taken **two analysts** full-time.[^3]

## Everyone is asking what you're doing about AI

Investors, customers and your own team all want to hear the AI plan, and each of them means something different by it.

> Property managers don't want an "AI Property Manager." They want to stop losing Sunday nights to spreadsheets.

The honest answer usually starts somewhere else. I start with the job, map the world it lives in, and write down what good looks like before anything is built. Then the model gets tested against it, with evaluations as release gates, not gut feel. The job comes from talking to the people who do it, **50–100 customer calls a year**.[^4]

## You want to try the new tools, and nobody has the time

A new model or agent comes out every week. Someone should find out which ones hold up on your real work, and everyone already has a full-time job.

I run them on my own work first. AbiertoClaw, a plugin I wrote, tests the free AI models **each day** with a real tool call and switches the agent to the one that passes.[^5] Brava's parent checkout on Stripe was built from scratch with AI coding agents. The tools on this site can be run by someone's AI agent, not just clicked.

## Your customers aren't in the room

The numbers say people do something odd, and nobody on the team has asked them why.

At AppFolio I sized how much rent was paid offline, from accounting receipts, then asked renters why: irregular gig pay, no bank account, a **15–30 minute** drive for certified funds, and slow ACH.[^6] Leadership funded teams to build rent-split payments and a digital certified-funds partnership.

## You need a first product person, not a department

There's product work every week and nobody whose job it is. The founder does it at night, between everything else.

At AppFolio I was the only product manager for the rental application, across **two codebases**.[^7] At Lively I went from data scientist to product manager as the company scaled. On Brava I design and ship end to end, from the first research call to the brand on the page. A block of hours each week can cover the work until you're ready to hire.

[^1]: AppFolio, on my résumé: "Rebuilt the rental application end to end. Submission rates went from 44% to 70%, and properties on the new flow fill vacancies 3.4 days faster than offline workflows."
[^2]: AppFolio, on my résumé: "Designed an assumed-success window and a tokenized failed-payment recovery path, reaching 99.1% application-payment completion."
[^3]: Oracle, on my résumé: "Automated a data pipeline that had taken two analysts full-time."
[^4]: Skills, on my résumé: "Product: jobs-to-be-done (50–100 customer calls a year), spec-driven development, tests/evals as release gates, OOUX"
[^5]: AbiertoClaw, on my résumé: "Each day it finds the best free model that passes a real tool-call test and switches Hermes to it, with a local model as the fallback."
[^6]: AppFolio, on my résumé: "On the payments team, sized offline rent from accounting receipts and interviewed renters on why they pay offline: irregular gig pay, unbanked renters driving 15–30 minutes for certified funds, 2–3 day ACH."
[^7]: AppFolio, on my résumé: "Sole PM for the rental application line, across two codebases."