A Palo Alto fintech pioneer had already started building AI support in-house. By building side by side with Notch, they turned tribal, employer-specific knowledge into dependable agents, and proved the system could handle their complexities at scale.
At a glance: Palo Alto · earned wage access pioneer · millions of users · hundreds of support reps · 1000+ payroll documents processed weekly
About EarnIn
EarnIn is a U.S.-based financial technology company, based in Palo Alto, that lets people access the wages they have already earned before payday. Through a mobile-first platform, EarnIn serves millions of users with earned wage access and financial wellness products, including Cash Out, Early Pay, and savings tools, all built to smooth cash flow and reduce financial stress. EarnIn operates in a highly regulated financial environment that demands secure, data-driven infrastructure that holds up at scale.
EarnIn's support organization matches that ambition. With hundreds of customer support reps and a strong internal engineering culture, the company had already started building AI into its support operation on its own. This is a team that knows its product deeply and takes real care with its customers. The question was never whether they were capable. It was how to move faster without lowering the bar they set for themselves.
The Challenge
Earned wage access is deceptively complex under the hood. The rules are different for every employer and every bank. Eligibility, pay schedules, and direct deposit behavior all vary, and a lot of the knowledge support reps rely on lives in experienced people's heads rather than a clean, documented process. It is tribal knowledge, built up over thousands of real cases.
When EarnIn began building AI support in-house, that complexity is exactly where it strained. Simple questions were fine. Complex queries and incomplete SOPs broke the in-house solution, and edge cases, like a blurry photo of a paycheck, an employer with an irregular biweekly schedule, and a mismatched direct deposit record, kept slipping through. The problem was never effort or organization. Turning years of employee-specific knowledge into scalable AI agents is a heavy lift, especially while running a live support operation. But that complexity is exactly where Notch thrives. It was that turning years of situational, employer-specific knowledge into something an AI agent can apply consistently is a hard problem to solve alone, especially while running a live support operation at the same time.
The Solution: building together with Notch
EarnIn did not have to choose between building and buying. They kept building, with Notch underneath and alongside them. Notch worked as a toolkit for building agents and ran side by side with EarnIn's own dev team and pipeline, which turned a stalled in-house effort into a fast, shared implementation, while EarnIn kept full ownership of what they were creating.
The work followed a clear loop:
- Analyze with deep research, going through real interactions to see how cases were actually handled and where the gaps were.
- Surface live suggestions and cover those gaps, as missing rules and exceptions appeared.
- Deploy new SOPs and knowledge, turning what had been tribal and undocumented into structured, dependable guidance the agents could apply every time.
EarnIn's team uses ADAM, the part of the Notch platform that speeds up building and improvement, for more than just standing up agents. They lean on it to map JSONs and data extracts, and to quickly create and connect API endpoints, both internal and external. Integration work that usually slows a build to a crawl moves at the same pace as everything else, which is a large part of why the implementation went so fast.
This is where the partnership mattered most. EarnIn's SOPs weren’t a mess. The team was well organized. But much of the real expertise was still situational and lived with experienced reps. Working together, we helped structure that knowledge, refined it, and turned it into clearer rules the agents and the human team could both rely on. Better SOPs came out of it. People stayed in the loop where judgment matters, and the operation underneath became cleaner, faster, and more consistent. What came out of it was not only an AI layer on top of the old process. It was a cleaner, better-documented operation underneath it.
Proof at scale: real-time payroll document ingestion
One of the clearest examples is payroll document ingestion. Verifying earnings means reading whatever a customer sends, and customers send everything. Mobile photos, blurry PDFs, screenshots of a banking app. Notch ingests these in real time, extracts the employer, pay schedule, and direct deposit data, validates it against the account record, and makes an automated eligibility decision.
The pipeline runs in five steps: ingest, segment, classify, extract, act. It handles this variability at a sustained rate of over 1,000 payroll documents per week, in a domain where every employer and every bank has different rules, and an eligibility decision directly affects whether a customer can access money they're counting on.
The Result
EarnIn kept ownership of its build and moved faster than it could have alone. By building on top of Notch and continuing their own initiatives, the team achieved higher-quality results more quickly, with agents who handle each interaction with genuine care. Their support organization now serves more customers, faster, while its people are freed to focus on higher-leverage work.
Why They Chose Notch
EarnIn chose Notch because it let them keep building rather than start over. Notch gave them a foundation to build on and a partner to build with. Our role was to help them see their patterns clearly, document them rigorously, and let AI amplify what was already working. That created a practice the whole team could trust and build on. It turned unclear, tribal knowledge into structured SOPs, and stood up dependable agents for a highly regulated banking and fintech environment, at the scale and complexity EarnIn’s business model requires. They did not trade control for speed. They got both.
"We needed an in-app concierge experience customers could truly rely on when accessing their hard-earned wages. The challenge was turning complex SOPs and tribal knowledge, with different rules for every employer and bank, into clear and dependable support. Notch broke it all down quickly and built agents that handled each interaction with real care. That has allowed our team to serve more customers, faster, while focusing on higher-leverage work."
Florencia Resnik, Vice President, Care and Agentic AI Operations, EarnIn
Bottom Line
EarnIn had already started building. Notch helped them finish faster and better by building alongside their team instead of replacing their work. Together with Notch, they turned tribal, employer-specific knowledge into structured SOPs and dependable agents, automated real-time payroll document verification at more than 1,000 documents a week, and freed the support organization to serve more customers and focus on higher-value work. With Notch, you don’t have to choose between building and buying. You can build on top of Notch and get there faster.
