Pomegra
Madrid-based Talky closed a €1.5M pre-seed round to build AI that learns each company's own accounting judgment, and it says the system already handles up to 93% of documents untouched.
Key Takeaways
- Talky raised €1.5M in pre-seed funding, announced October 29, 2026, with no lead investor or valuation disclosed.
- The platform says it processes 86% to 93% of documents without human intervention.
- Talky reports 1.8M documents processed and over €2B in client spend, with a team of about 10.
Lead
Talky, a Madrid startup founded in 2025, announced on October 29, 2026 that it raised €1.5M in a pre-seed round. The deal was signed in September 2026. The venture funds HelloWorld, Silence VC and Motive Force took part, alongside three angels: Sebastián Gunningham, CEO of Remitly; Paul Golbaum, CEO of Lightbase and former CTO of Seedtag; and Javier Morenés, an early investor in Onum. The company did not name a lead investor or disclose a valuation.
What Does Talky Actually Build?
Talky builds an AI platform that automates accounting and operational finance for the finance departments of large companies. It studies a company's historical ledger entries and the decisions its finance team and auditors have made. It then applies those patterns to new documents, posts invoices to the company's ERP, and attaches an audit-ready justification to each entry.
The pitch rests on a narrow observation. Accounting rules are public, but each company applies them through house conventions, exceptions and precedents that rarely appear in a manual. CEO Jorge Beamonte put it this way: "Each company has its own way of doing accounting, and it is almost never written down." Talky's software tries to infer those unwritten criteria, route difficult cases through them, and keep a traceable record of why it chose what it chose.
The product competes with a crowded field. Incumbent ERP modules, invoice-capture vendors and a wave of AI-first finance startups all sell some version of document automation. Talky's differentiator is the learning step on a customer's own history, not the capture step.
How Strong Are the Numbers?
The company reports that it processes 86% to 93% of documents without human intervention. It also says it has handled more than 1.8M documents representing over €2B in client spending. Those figures come from Talky itself and have not been independently audited.
The range matters as much as the ceiling. A straight-through rate that swings by seven points suggests results vary by customer and document type, which is typical when accuracy depends on how consistent a client's past bookkeeping was. The remaining 7% to 14% of documents are the hard cases, and in accounting those tend to carry most of the risk. Talky has not named its customers.
Who Is Behind the Company?
Beamonte, 25, is CEO, and Maximilian Letón, 23, is chief operating officer. The team numbers about 10, nine of them with software engineering and mathematics backgrounds. Neither founder's earlier track record is documented in the announcement, so the angel list carries part of the credibility load: operators from a payments company, an adtech company and a Spanish observability startup.
The absence of a named lead is common at pre-seed, where funds and angels often split a small cheque across several names. It does mean the round gives little signal about who sets terms or sits on the board.
What Will the Money Fund?
Talky plans to spend the €1.5M on product and technical infrastructure, on deepening work with existing clients, and on international expansion. At a team of 10, the round probably covers hiring and compute for roughly the next 12 to 18 months, though the company has not stated a runway.
The expansion goal is the more ambitious item. Accounting practice differs by country in tax treatment, chart-of-accounts conventions and audit expectations. A model trained on one company's habits has to be rebuilt, or at least retuned, for each new jurisdiction and each new ERP.
Why Does This Round Matter in European Finance Automation?
The round matters because it shows investors still fund narrow, evidence-led AI finance tools at the earliest stage. Talky joins a run of European back-office automation rounds, including Spain's Centinel (€1.2M), Germany's mika (€6M) and Belgium's Ravical (€17M). Against those, €1.5M is small, and it buys a proof-of-concept extension rather than a land grab.
Market forecasts add momentum but little precision. One estimate puts the global AI-for-accounting market at $54.2B by 2034, a 25.9% compound annual growth rate. Such projections are broad and vendor-driven, and they say less about Talky's prospects than its customer retention will.
What Comes Next for Talky?
The next test is whether the reported automation rates hold as Talky adds customers whose books are messier and whose auditors are stricter. A seed round, probably sized at several times this one, would likely depend on named enterprise clients and evidence of expansion outside Spain. If straight-through rates fall as the customer base widens, the "learns your judgment" claim weakens into ordinary document capture.
Incumbent ERP and spend-management vendors are adding their own AI features, which compresses the window for a small team. Talky's counter is depth: an audit trail per entry is hard to bolt on later and matters to the finance chiefs who sign off on automation.
Outlook
Talky has funding, a stated automation rate of 86% to 93%, and a thesis that company-specific accounting judgment can be learned from history. It lacks a disclosed valuation, a named lead and public customers. Over the next year, the evidence that matters is whether its automation rate survives new clients and new countries.
