There is a room in almost every large company where someone opens a PDF, squints at a supplier's invoice, and types the numbers into a system by hand. Then they do it again. And again. In a multinational with hundreds of sites, that room never really closes. Cedalio, a Y Combinator company from the Summer 2023 batch, was built to empty it - to hand the reading, matching, and coding of invoices to a set of AI agents and leave the humans to decide only what the machines flag as strange.
The company's current pitch is unusually plain for a startup: "the AI layer your ERP is missing." Cedalio does not try to replace SAP or Oracle or whatever ledger a finance team already runs. It sits in front of that system and cleans up the mess before it arrives - invoices, purchase orders, receipts, and utility bills that show up in a dozen formats and several languages. The agents capture documents from email, vendor portals, and shared drives, pull out the data with vision models rather than old-style optical character recognition, run a three-way match against the purchase order and the goods receipt, and pass anything that does not reconcile to a person.
Figures as published by Cedalio; treat vendor-reported metrics as directional.
01 / The Work Nobody WantsWhat Cedalio actually does
Accounts payable is not a glamorous corner of software. It is also where a company's money physically leaves the building, which is exactly why errors there are expensive and fraud there is quiet. Cedalio's ten specialized agents run the full cycle overnight: one reads, one matches, one audits, one codes each entry to the right account. By morning the finance team is looking at a short list of exceptions instead of a tall stack of paper. The company frames this as a shift in what people do all day - from manual entry to governance and oversight.
Two features do a lot of the heavy lifting. The first is duplicate and anomaly detection, which catches the invoice a vendor "resent" that would otherwise get paid twice, and the tariff on a utility bill that quietly crept up. The second is Latin American tax compliance, including the withholding calculations that make LATAM accounts payable genuinely painful and that most horizontal tools treat as an afterthought.
It helps to picture the before and after. Before, a payables clerk logs into three vendor portals, downloads a folder of PDFs, opens each one, copies the invoice number and totals into the ERP, then hunts through a separate system for the matching purchase order to confirm the company actually ordered what it is being billed for. A single exception - a quantity that is off by two units, a tax line that does not add up - can stall a payment for days. After, the same documents arrive, get read and matched automatically, and only the genuine mismatches surface. The clerk's job stops being data entry and starts being judgment.
"Say goodbye to manual data wrangling and hello to real-time, centralized information."Cedalio, product launch
02 / A Company That Kept Its ThesisThree pivots, one idea
Here is the part that makes Cedalio interesting rather than just useful. It has changed shape three times. It started in 2022 as a blockchain-verifiable database - a GraphQL-accessible, decentralized store whose whole promise was that any piece of data could be traced back to its origin and proven untampered. When YC's Summer 2023 batch rolled around, the company had followed its users into sustainability, launching a platform that pulled energy, water, gas, and emissions data off utility bills and turned it into auditable ESG reports. Then it followed them again, into the broader problem those same bills belonged to: the finance department.
Most pivots throw away the old thing. Cedalio's did not. The audit trail that justified the blockchain in 2022 is the same audit trail a CFO wants on every invoice in 2025. The founders appear to have treated "make data trustworthy and traceable" as the actual product and the packaging - decentralized database, sustainability platform, finance agents - as negotiable. That is a subtle discipline. It is easy to fall in love with a technology. It is harder to notice that customers keep asking for the same underlying thing and to follow that instead.
The evolution also explains why the utility-bill capability never disappeared. It was the bridge. Bills are documents that finance, energy, and operations teams all fight over, so a tool that could read them cleanly had a natural path out of sustainability reporting and into accounts payable, where the same normalization work is worth far more.
There is a quieter lesson in the sequence for any founder watching. Cedalio did not brainstorm its way from blockchain to invoices; it moved because customers kept dragging it toward the harder, more valuable version of the same problem. The blockchain users wanted proof their data was real. The sustainability users wanted that proof applied to bills. The finance users wanted the whole reading-and-matching machine, not just the ledger. Each step was a shorter leap than it looks from the outside, because the underlying job - turn messy documents into trustworthy data - stayed put while the buyer got richer and the pain got sharper.
03 / Who Is BuyingThe customers
Cedalio's client list leans heavily Latin American and heavily enterprise - the kind of multi-site organizations where invoice volume is high and the local tax rules are unforgiving. The named customers span banking, beverages, insurance, retail, and agribusiness, which tracks with a product whose value scales with document chaos rather than any single industry.
The go-to-market is deliberately narrow at the start. A customer picks one workflow - usually accounts payable or utility-bill processing - and Cedalio gets it live in one to two weeks. That short runway matters. The faster a finance team sees invoices clearing without a human keying them, the sooner it trusts the agents with the next workflow. It is land-and-expand, but the land is small enough to say yes to.
The buyer profile is worth noticing too. These are not scrappy startups looking for a slick app; they are established institutions with real audit requirements, real regulators, and real consequences for a payment that goes out wrong. That is a demanding first market, and it is a telling one. A company that can satisfy a bank's compliance team on accuracy and traceability has cleared a bar most invoice tools never have to face. It also means adoption is slower and stickier - harder to win, harder to lose once won.
"You don't need to understand the blockchain to benefit from it."Luciana Reznik, CEO & co-founder
04 / The Team Behind ItFounders with an exit already
Cedalio is not a first-timers' company. CEO Luciana Reznik, CTO Guido Marucci Blas, and chief product officer Nico Magni previously worked together at Wolox, an Argentine software firm that Accenture acquired in January 2021. Reznik, a software engineer from Buenos Aires who has been based in the Bay Area since 2018, studied at the Instituto Tecnologico de Buenos Aires and did executive training at Stanford's business school. The team is small - about six people - and split across hemispheres, with Reznik in San Francisco and Magni working from Patagonia.
The founders' pedigree shows up in the cap table. Alongside Y Combinator, Cedalio's early backers include Guillermo Rauch of Vercel, Esteban Ordano of Decentraland, and Demian Brener of OpenZeppelin - a roster drawn from web infrastructure and crypto, which fits a company that started on a blockchain database. Reported early funding sits around $1.5 million across a handful of rounds, though the company's YC-era seed is modest by design.
05 / The FieldHow it stacks up
Accounts payable automation is a crowded category. In the United States, Ramp, Bill.com, Tipalti, Stampli, Vic.ai, and Rossum all chase versions of the same promise. Cedalio's argument is not that it invented invoice reading; it is that the horizontal players treat Latin American tax, withholding, and multi-language, multi-site bills as edge cases, while Cedalio treats them as the point.
| Dimension | Typical horizontal AP tool | Cedalio |
|---|---|---|
| Extraction | OCR-based | Vision models, ~99% accuracy |
| LATAM tax & withholding | Limited / add-on | Built into the workflow |
| Utility & service bills | Rare | First-class, tariff validation |
| Model | Software you operate | Agents that run overnight |
| Audit trail | Logging | Verifiability as founding thesis |
Where the company sits in the market, then, is a specific seam: enterprise finance automation for organizations whose paperwork is too regional and too messy for the one-size tools, delivered as supervised agents rather than as another dashboard someone has to learn. It is a smaller opening than "automate all of finance," which is probably why it is a real one.
06 / The BusinessHow the money works
Cedalio sells B2B software, and it sells it on arithmetic. The company pitches a drop in the cost of processing an invoice from roughly $10 to about $1, an ROI it puts near twelve times, and a payback period under three months. Those are vendor numbers and worth reading as directional, but they point at a clear buyer logic: automation that pays for itself inside a quarter is an easy internal sell, and it gives the sales team a reason to start with one workflow and grow from there.
The expertise underneath is the reason the arithmetic holds. Reading a real-world invoice is deceptively hard - stamps, handwriting, tables that wrap, three languages on one page - and the founders bet on vision models over OCR precisely because that is closer to how a person reads than how a scanner does. Pair that with a decade of enterprise-software experience from the Wolox years and a founding obsession with verifiable data, and you get a company that is credible in a category where accuracy is not a nice-to-have but the whole job.
If there is a risk in the story, it is the same one that comes with any tightly-scoped wedge: the horizontal giants can eventually add LATAM depth, and enterprise finance is slow to switch tools. Cedalio's answer, so far, is speed of deployment and a product that keeps proving itself one workflow at a time. For a six-person team that has already followed its customers across three different products, that patience looks less like indecision and more like the plan.