The $15-an-Hour Expert: what YC wants and how to validate it
Original request by Brad Flora, Y Combinator. Source: YC Requests for Startups.
The request in plain English
Instead of AI that replaces workers, YC wants AI that upgrades them: software that lets someone without a professional credential do work that used to need a specialist. The target is any field where the credential, not the work itself, is the bottleneck.
What YC's request says
- Example: a customer service rep who can advise on M&A, supply chain, or global trade with AI support.
- Target areas named: benefits, taxes, insurance claims, medical billing, and immigration paperwork.
- The framing is building a new middle class of expert jobs, not cutting headcount.
Why now
A lot of expert work is pattern-matching against rules: which code applies, which form is needed, which exception covers this case. Models are now good enough to carry that knowledge, explain it, and show their reasoning, which means the human in the loop no longer needs years of training to get the right answer.
At the same time, many of these fields have real shortages. There are not enough people to process claims, appeals, and paperwork at the speed customers expect. That makes the pitch to a buyer simple: more throughput from the people you already have, or new hires who become productive in weeks instead of years.
Who pays
The cleanest buyers are businesses that already employ large teams doing semi-expert work and struggle to hire: medical billing companies, third-party claims administrators, tax preparation chains, benefits brokers, and immigration law practices with high paralegal load.
Selling to the workers directly is harder. The money sits with the employer, who measures throughput, error rates, and time to productivity. Build your metrics around those numbers from day one.
Three startup ideas to start from
Medical billing copilot
Guide a non-certified biller through coding, denials, and appeals, with every suggestion tied to the payer rule behind it.
Benefits navigator for HR teams
Let a generalist HR coordinator answer complex leave, disability, and benefits questions that currently get escalated to specialists.
Immigration paperwork assistant
Help paralegals assemble and check filings, flagging missing evidence before a case is submitted.
How to validate it before you build
- Pick one workflow and measure how long a credentialed expert takes versus a trained generalist today. That gap is your product.
- Sit with five teams doing the work and count escalations per day. Escalations are the clearest sign the credential is the bottleneck.
- Run a pilot where generalists use your tool on real cases and an expert reviews the output. Track error rates, not just speed.
- Find out who is liable when the answer is wrong, before you sell. In regulated fields, that question decides the deal.
The founder trap
Building a general 'AI expert' instead of owning one narrow workflow end to end. Buyers trust tools that are right about one thing every time, not tools that are roughly right about everything.
IdeaCheck's take
Strong market pull and clear ROI, but liability and accuracy carry the risk. Start in a workflow where mistakes are caught by review before they cost anyone money.
Test your version
Check your idea against Hacker News
Building something in ai and work? See what 28,000 Hacker News threads say before you commit.
Your first evaluation is free with Google sign-in.