Six months in, and Rex is making a claim that would have seemed fanciful just a few years ago: its AI agents are chasing down invoices, logging into supplier portals, and hauling in overdue payments for pre-IPO tech companies—all without a human touching the keyboard.
The London- and San Francisco-based startup says it's processing north of $500 million in receivables—a company-reported figure that lacks independent verification—through software that reads emails, submits invoices to platforms like SAP Ariba, and follows up on late-paying Fortune 100 customers. For a company that graduated from Y Combinator barely months ago, those are eye-opening figures. Perhaps more eye-opening is the premise: that the painstaking, personality-heavy work of accounts receivable—a function companies have outsourced to armies of humans for decades—can be handed off to autonomous agents.
Rex positions itself less as a software vendor and more as an "AI operations partner," a distinction that may or may not hold up under scrutiny but signals where the founders think the market is headed. The company's pitch centers on a familiar pain point for any finance chief: days sales outstanding, that gnawing metric that tracks how long it takes to convert an invoice into actual cash. Rex claims it can shave an average of 12 days off that cycle, cutting manual follow-up from weeks to hours.
If true, it's the kind of improvement that gets a CFO's attention—especially one watching working capital in a tighter funding environment.
The Mechanics: Supervised Autonomy, Mostly
What does Rex actually do? The platform integrates with the usual suspects in enterprise finance—Oracle NetSuite, SAP, Microsoft Dynamics, Workday, Salesforce—and maps a company's order-to-cash workflow. From there, agents take over the repetitive, high-volume tasks: retrieving purchase orders buried in supplier portals, matching remittance advices to invoices, sending dunning emails, reconciling payments. The system logs every action in an audit trail and escalates edge cases to humans when things get messy.
The company reports auto-triage rates above 90% and average response times around 30 minutes, though those numbers come from Rex itself and lack third-party validation. Still, co-founders Merlin Kafka and Lewis Blackwood—both alumni of Sequence, the a16z-backed fintech—insist the platform is already handling material volumes for real customers.
Their main proof point? Synthesia, the AI video generation platform, which Rex lists as its sole publicly disclosed customer as of early July. According to a case study on Rex's site, Synthesia now routes roughly 2,000 AR cases per month through the platform. In the first four months, Rex's agents processed around 385 remittance advices and promises to pay, synced 500 purchase orders, and submitted more than 100 invoices monthly across various portals. The agents executed over 22,000 autonomous actions and retrieved more than $40 million in purchase orders sitting idle in supplier systems.
It's positioned as proof of concept at scale. The case study includes quotes from Synthesia's CFO and AR team, though there's no explicit publish date beyond a 2026 timestamp—a small detail that underscores just how early-stage this all is.
A Compressed Timeline
Rex's trajectory has been swift, even by startup standards. The company was incorporated in January. By April, it had won Vercel's AI Accelerator—beating out some 2,500 applicants—and secured investment from Vercel Ventures. In June, Rex claims to have joined the NVIDIA and Deloitte "Adopt 100" program, a cohort launched during London Tech Week to accelerate enterprise AI adoption, though this participation has not been confirmed by official program announcements.
The founders bring credible resumes. Kafka was chief product officer at Sequence and earlier worked as a product manager at Tessian, the Sequoia-backed email security startup that Proofpoint acquired in 2023. Blackwood was a founding engineer at Sequence and, before that, an early hire at GoCardless. He also co-founded Personably, a GFC-backed venture, before the Sequence stint.
But the scale of the operation raises questions. Rex's Y Combinator profile listed the team size as two as of June—lean, to say the least, for a company claiming to manage hundreds of millions in receivables. LinkedIn shows the headcount somewhere between two and ten employees as of July, a discrepancy that suggests either aggressive automation of their own operations, an ongoing hiring wave, or a very early-stage deployment footprint.
Entering a Crowded, Noisy Market
Rex isn't exactly breaking into virgin territory. The order-to-cash automation space is already thick with incumbents and a fresh wave of AI-first challengers, all of whom have been racing to rebrand legacy software as "agentic AI."
HighRadius has been trumpeting its agentic capabilities since 2025, claiming to deploy more than 180 agents across its platform. Sidetrade announced an AI cash collection agent in May 2025. Emagia launched "Gia AlphaCash" in March. Billtrust introduced agentic voice-over-IP for collections in January. The terminology is everywhere; the differentiation, less clear.
Rex's counter-positioning is interesting, though. Rather than selling software with a dashboard, the company frames itself as a service layer—closer to replacing a business process outsourcer than installing another finance tool. The company cites the $60 billion outsourced O2C function as its addressable market in its YC materials. Industry research from Grand View Research pegs the dedicated O2C BPO market at around $37 billion in 2025, with a broader finance and accounting BPO market near $76 billion.
That framing—agents as labor replacement, not just software—could resonate with finance leaders weary of maintaining sprawling offshore teams. Or it could run headfirst into the same skepticism that has dogged other attempts to automate inherently relationship-heavy processes.
The Enterprise Readiness Question
For all the momentum, there are still foundational gaps. Rex's website lists SOC 2 Type II compliance as "pending" as of early July—a detail that matters considerably when pitching Fortune 100 enterprises on handing over invoice data and customer payment records. The company does advertise GDPR compliance, EU data residency, AES-256 encryption, enterprise single sign-on, and a pledge not to train models on customer data. But pending certifications signal a company still building its enterprise scaffolding.
Pricing is tied to outcomes and service-level agreements, according to the site, with Rex promising return on investment within a month. That timeline feels aggressive—though it likely depends on the scope of deployment and how tangled a given AR operation is to begin with.
The company also offers "Compass by Rex," a two-to-four-week diagnostic engagement meant to surface cash-delaying exceptions and sequence agents to tackle root causes. It's part sales tool, part operational philosophy: deploy agents where the problems cluster, not just where the automation is easiest.
An Open Question
Whether Rex can scale from Y Combinator demo day to credible market challenger remains an open question. The startup is betting that finance leaders are ready to believe agents can handle work they've outsourced to humans for generations—work that often requires judgment calls, relationship management, and the occasional tough phone call with a late-paying customer.
The early numbers from Synthesia suggest Rex isn't just vaporware. But one customer, however impressive, doesn't make a category. The real test will come when Rex tries to move upmarket, when the contracts get bigger and the compliance requirements steeper, and when the competition—well-funded, entrenched, and now equally AI-obsessed—decides whether to ignore the newcomer or crush it.
For now, Rex is placing a bet that the future of accounts receivable looks less like a team in Bangalore and more like a swarm of autonomous agents working around the clock. If they're right, it's a sizable market up for grabs. If they're wrong, well—there's always next batch.
