The Software Kept the Records. It Never Did the Work.

Every finance team in the world runs on a quiet fiction: that the software is doing the work. It isn’t. The ERP holds the record. The billing system holds the invoice. The bank holds the cash. In between, a person opens their inbox at eight in the morning and starts chasing. They email a vendor about a short payment. They match a wire against four invoices and a credit memo. They upload a remittance PDF, correct an entry, follow up, follow up again. The system of record recorded. It never did anything.

McKinsey puts as much as $4.4 trillion a year of expert work in play as AI moves from drafting to doing. Most of the attention has gone to the glamorous end of that number: lawyers, radiologists, engineers. We think the first trillion comes from somewhere far less interesting and far more mechanical, the back office of every mid-market company on earth, where the work was never knowledge work at all. It was clerical work wearing a knowledge-work salary.

Founders who were the exception queue

Tom Dowling, Geoff Segal and Aaron Coppa did not read a market map and pick a category. They lived inside this problem for four years.

Their first company, TaxProper, went through Y Combinator’s S19 batch and became the property tax platform for institutional real estate. At scale, it moved more than a billion dollars a year in property tax payments on behalf of its customers, until Opendoor acquired the business in 2022. That is a real exit and a real operating record. It is not the part of the story that made us write the check.

The part that made us write the check is what happened inside their own company while it was growing. Their accounting team drowned. Invoicing, reconciliation, customer questions, exception handling: the volume expanded faster than headcount could reasonably follow, and every tool they bought kept better records without removing a single hour of work.

Their conclusion, which is the entire thesis of the new company: accounting teams do not need another dashboard. They need teammates.

FullSeam (YC W26) builds agents that log into the accounting tools a company already runs and perform the tasks AR and AP teams do by hand: communicating with vendors and customers, recording payments, updating financial records, running cash application and reconciliation. No migration. No new system of record. The agent works inside the stack the controller already trusts.

Dowling’s route to that desk is its own argument for our thesis. A Rhodes Scholar with an Oxford MSc who advised the Mayor of Chicago and was appointed by Illinois’s governor to oversee a state agency spending $700 million a year, and then went to Y Combinator, twice. That is not a linear career. It is the profile of someone who has repeatedly walked into systems that were failing and been handed the keys.

Why this is a Sunshine Lake shape

Extreme origins produce specific knowledge. Not “AI for finance.” The exception queue. The remittance PDF. The vendor who will only confirm by phone. You cannot reach that resolution from a market map. You reach it by having been the person on the phone.

The buyer is already convinced. Enterprise buying behaviour outside the glamour categories inflected hard in the last three to five years. Operators who would previously only engage a startup through an innovation programme now buy with the main checkbook, because their competitors are compounding advantage and they can feel it. The category does not need evangelising. It needs serving.

The capital requirement is small and the revenue is early. FullSeam needs a design partner, a working agent, and a controller willing to hand over the AR inbox. It does not need a decade of pre-revenue physics. Companies shaped this way reach commercial proof before they need permission from the capital markets to continue, which is the only form of independence that survives a downturn.

The world’s accounting departments are roughly the least romantic market available. That is precisely why the software that finally does the work will be worth what it will be worth.