The manual busywork of finance operations has long been a punishing economics problem disguised as a workflow challenge. Aaron Coppa, Thomas Dowling, and Geoff Segal spent enough years watching it up close—the invoice chasing, the three-way matching, the exception handling that consumed hours while generating precisely zero strategic insight—that when Opendoor acquired their startup TaxProper in 2022, they left with something more valuable than the $10 million check. They left with conviction.
Now the trio is back with FullSeam, emerging from Y Combinator with a product that tries to solve what dozens of finance automation vendors have promised but rarely delivered: AI agents that don't just surface problems or suggest fixes, but actually complete the work themselves.
The distinction matters more than it sounds. Finance teams have spent the better part of a decade integrating workflow tools that mostly shuffled tasks around, flagged anomalies, or required elaborate configuration before doing anything useful. FullSeam's pitch is different—perhaps uncomfortably so for incumbents. Its agents plug directly into a company's existing software stack (QuickBooks, NetSuite, Stripe, HubSpot, Salesforce) and operate autonomously within guardrails the finance team sets.
No migration. No rip-and-replace. Just agents that log in where humans used to and start working.
How autonomous looks in practice
What does "autonomous" actually mean here? The AR agents chase outstanding payments, parse customer emails for blockers like missing purchase orders, update invoices in real time across systems, then draft follow-up messages. On the payables side, agents ingest invoices from email and vendor portals, match them against purchase orders and contracts, code transactions to the general ledger with confidence scores attached, and execute three-way matching before anything hits the ERP.
Cash application—matching payments to invoices—gets handled with what FullSeam calls "explainable rationale," a nod to the audibility concerns that follow any automation into the CFO's office. Bank reconciliation follows a similar pattern: flag variances, surface exceptions, log everything. When an agent encounters something it can't resolve or hits a threshold requiring human judgment, it routes the issue up with full context rather than guessing or stalling out silently.
"We're positioning these as AI operators, not workflows," Dowling said in the company's YC launch materials. The language is deliberate. FullSeam isn't trying to be middleware or another layer in the data stack; it's positioning itself as connective tissue that sits above existing systems.
Whether that's a defensible moat or just good positioning remains an open question.
Early returns (with caveats about scale)
Two early customers offer a glimpse of what the product does when it works. Deferred, a referral management platform, now processes all referral payouts automatically—saving north of 10 hours monthly and moving payouts to real-time. CEO Judd Schoenholtz said the system is set to scale to more than 10,000 referral partners without adding headcount, though it's worth noting Deferred is still early in that scaling curve.
Belfry Software reports 95% of invoices fully validated by AI before the finance team sees them, a 40–60% drop in billing disputes, and about 10 hours returned to finance each month. "This has fundamentally changed how we operate," CEO Jordan Wallach said—a quote that reads like typical vendor case study language but tracks with the time savings other automation tools have delivered when they actually stick.
The company claims that 60% of accounting time is manual work, a figure that feels plausible if not rigorously sourced. What's clearer is that FullSeam isn't alone in chasing this market. Serrala launched its own agentic AI finance platform earlier this year, targeting the Office of the CFO with autonomous agents across AR, AP, and payments. Oracle followed with agentic AI for retail banking collections and compliance. The category is heating up fast, which means FullSeam's window to establish itself is narrowing.
Speed, guardrails, and the trust problem
The founders emphasize deployment speed—days instead of months, they say—by connecting to tools finance teams already use rather than forcing an infrastructure overhaul. Agents operate within rules the finance team configures: approval thresholds, new vendor review requirements, general ledger coding confidence minimums. It's a control framework designed to address the trust gap that still dogs AI in finance, where errors don't just annoy users but create audit exposure.
Integration logos on FullSeam's website span QuickBooks, NetSuite, Xero, Stripe, Bill, Sage, PandaDoc, and HubSpot, though the company hasn't published detailed certification documentation or security attestations publicly yet. The Terms page lists standard SaaS liability language under Delaware law—table stakes for an early-stage startup, but something larger customers will scrutinize as conversations get serious.
The team's second act
All three founders worked together at TaxProper, where they built tax automation technology before Opendoor acquired the company in late 2022. Dowling, a Rhodes Scholar who studied at the University of Illinois and Oxford, ran TaxProper as CEO. The current team is small—three people, working with YC partner Brad Flora—which means the product is either elegantly simple or about to hit scaling challenges as customers come onboard.
FullSeam registered as a Delaware corporation with a New York foreign filing late last year and went live with its website and first customers shortly after. The YC launch came in mid-February, timed around the standard founder ask: introductions to finance teams still drowning in manual AR and AP work.
The target customer is clear enough—finance and accounting teams running multi-system contract-to-cash operations with high exception volumes. These are environments where manual work scales linearly with growth, and where ten hours saved per month compounds quickly as the business expands. It's also where legacy vendors have entrenched relationships and switching costs that FullSeam will need to overcome with more than a slick demo.
The harder question
The challenge for FullSeam isn't technical capability—early results suggest the product works—but whether finance teams are ready to hand off execution to AI agents, even auditable ones. Trust in automation has always lagged functionality in finance operations, a gap born from painful experiences with tools that promised autonomy but delivered fragility. FullSeam inherits that skepticism whether it deserves it or not.
Still, the founders have been here before. They built and sold one finance automation business already, which means they understand the sales cycles, the procurement hurdles, and the feature requests that matter versus the ones that don't. That experience counts for something, maybe more than the $10 million exit itself. The question now is whether the market has finally caught up to what they're building—or whether FullSeam is still just early enough to be interesting and too early to be inevitable.
