The onboarding bottleneck that automated itself away

Every new customer arrived with their own data, in their own format. Bringing them on meant an engineer sitting down and mapping that customer’s files into the company’s internal system — by hand, from scratch, every single time. It took weeks of engineering time per customer.

It was slow and error-prone, but the real cost was subtler: it put growth on a leash. The more customers the business won, the more engineering it burned just to onboard them. Success made the bottleneck worse.

What I built

As lead architect, I designed and built an integration platform that automated the file-mapping. Instead of an engineer translating each new customer’s data by hand, the platform did the translation itself — taking customer files in whatever shape they arrived and mapping them into the internal model automatically.

Onboarding stopped being an engineering project and became something that mostly just happened.

The result

The work that had been burning weeks per customer now ran automatically, freeing the engineering team to build product instead of doing data translation by hand — an estimated $10 million a year in recovered velocity and labor. And because onboarding no longer scaled with headcount, growth stopped being throttled by it.

Why this matters for a business like yours

Most businesses have a smaller version of this exact problem: data arriving in one shape that someone re-keys into another, over and over — invoices into the accounting system, orders into the CRM, spreadsheets into a database. That re-keying is usually the first thing worth automating. This was that same problem at fintech scale, but the principle is identical whether it’s customer files or the ones landing in your inbox this week.