Madrid's Talky raises €1.5M to teach AI unwritten accounting rules

Madrid-based Talky Labs has raised €1.5 million in a pre-seed round to build AI that learns how each company actually keeps its books — criteria its founders say are almost never written down.

The round, closed in September 2026, drew participation from venture firms HelloWorld, Silence VC, Motive Force and Aera Capital. Angel investors included Sebastián Gunningham; Paul Golbaum, CEO of Lightbase and former CTO of Seedtag; and Javier Morenés, among others. Talky plans to spend the money on hiring, product development and technical infrastructure.

The company is small by any measure: nine employees. Yet it reports having processed more than 1.8 million documents and over €2 billion in client expenditure through its platform — roughly €220 million in annualized spending handled per head, if the figures hold. Those numbers are self-reported; Talky did not disclose revenue or client names.

What does the technology actually do?

Talky targets a gap that conventional automation leaves open. Standard tools handle routine transactions, but unusual invoices and exceptions still require manual work, causing delays and errors. The relevant information sits scattered across ERP systems, spreadsheets, emails, contracts and the heads of individual employees.

The platform takes a different route from rule-based software. Its AI learns from a company's historical accounting records and the expertise of its finance teams, identifies how previous transactions were handled, extracts the accounting criteria and applies them to new operations. When the system meets an unfamiliar case, it flags the right employee to consult rather than guessing.

According to the startup, the technology processes between 86% and 93% of accounting documents without human intervention. In plain terms, up to roughly one document in seven still needs a person. The platform reads invoices, suggests accounting entries, reviews tax information and reconciles bank transactions before writing results into the client's ERP system.

Who is behind the company?

Jorge Beamonte, CEO, and Maximilian Letón, 23, COO, founded Talky in 2025. The two met at a networking event two years earlier — a moment Letón described in unusually personal terms for a funding announcement.

"Two years ago, I met my co-founder Jorge Beamonte at a networking event. That night, I called my mom and told her: 'Mom, today is the most important day of my life. Everything is about to change,'" Letón said in a public statement.

Beamonte framed the product pitch around institutional knowledge. "Every company has its own way of doing accounting, and it's almost never written down. Talky learns those criteria, applies them to the hard cases and turns them into information for better decisions," he said.

Is this a one-off deal or a wave?

Neither. Talky's round sits inside a run of 2026 European investments in AI-powered accounting and finance automation, and the pattern is worth a skeptic's attention:

• Belgium's Dytto raised €1.5 million for an AI assistant for accountants
• Denmark's Plinto secured €2.9 million for automated financial controls
• Germany's mika raised €6 million for AI accounting and tax services
• The UK's Zalos obtained €3.1 million for enterprise finance automation
• Spain's Dolfin, based in Barcelona, took €2.1 million for adjacent financial workflows

Together, these rounds total approximately €66.8 million including Talky's investment — a figure that implies considerably larger undisclosed rounds elsewhere in the set, since the six named deals sum to well under €20 million. The source does not break down the remainder.

What comes next?

Talky says it will use the funding to expand among large enterprises and financial services providers handling high transaction volumes. The founders call recruitment the central priority, planning to grow the team while developing the platform with existing customers.

A pre-seed round buys runway, not proof. The test will be whether Talky's claimed 86-93% automation rate survives the messy exceptions of large corporate clients — the very cases it says it was built to solve.