Porters

The models replace a manual outsourcing operation, which is precisely what the company positions against. Its own framing is that institutions currently rely heavily on outsourcing and that banking operations have remained stubbornly manual despite decades of digitalization. The proposition is described as AI native systems replacing legacy tooling with autonomous workflows, and as an ambition to run entire services end to end through models rather than to assist people doing them. The founding technologist has a doctorate from a leading technical university and built applied machine learning systems before this.

The design position is stated carefully. Workflows are described as autonomous but human in the loop, capable of safely handling regulated tasks while keeping compliance safeguards and traceability intact. Naming traceability alongside human involvement matters because these processes end in a court file, and an institution asked why an account was seized relies on that trail to reconstruct the decision.

Against that is the stated direction of travel, an ambition to run entire services end to end through models and a benefit framed as reducing manual oversight. Nothing defines where the human sits, what triggers escalation, or which steps remain supervised as the platform matures.

One control is named rather than implied. Traceability is stated as a property preserved alongside compliance safeguards, which for a court driven process is the right one, since the requirement is to show what was done and why. No accuracy or error rate is published, nor any validation result or model documentation, which is expected for a company at this stage. What is absent is any measure of correctness. For automated seizure handling the operative figures would be how often the right account is identified and how often a deadline is met.

The company was founded in 2025 and has six people moving to twelve, so there is no operational record yet and no customer is named. Two things carry weight for a business this young. It won early stage startup of the year at its country's principal financial technology awards in June 2026, where the jury stated it addresses a clear need and is particularly well positioned to execute, and every finalist that year was an artificial intelligence company.

And the pre seed round drew four operating fintech founders as angels alongside a European venture firm with more than 2.5 billion euros under management and a well known accelerator, which is capital from people who have run the operations this product targets.

Many banks and fintechs running the same statutory processes are the intended customers. The operational learning the platform accumulates (which court formats work, how a given authority responds, where cases stall) is directly valuable across that customer base, and those customers compete with one another. No data boundary statement is published. Nothing states what is retained from one institution's caseload, whether it informs another's, or how customer level records are separated within a shared platform.

The payload is highly sensitive. Account seizure and insolvency files record that a named customer is subject to court enforcement or is insolvent. That is among the most consequential facts a bank holds about a person, and one that materially affects them if it leaks or persists after resolution. There is no public data protection agreement or retention schedule, and no subprocessor list or deletion commitment. Nothing published states how those records are held, how long they survive the closure of a case, or how they are separated between institutions.

An outsourcing arrangement handling regulated processes on customer accounts triggers third party risk assessment and, in several European jurisdictions, supervisory notification before the arrangement can begin. That makes security assurance the gate this product faces earliest. There is no public attestation or certification, and no trust center or enumerated framework, which is unsurprising for a company months past its first round.

Compliance readiness is claimed throughout, with the workflows described as regulated and the platform as compliance ready, yet no statute, court procedure or supervisor is named. That gap matters for this product because the process is the regulation. Garnishment and insolvency handling exist only as creatures of national debt enforcement law, with prescribed timescales, notification duties and priority rules, and fines follow directly from getting them wrong. Naming the enforcement regime it automates would be the single most informative disclosure available to it.

No credit decision is involved, so the exposure is severity rather than selection. The people whose accounts these processes touch are at their most financially vulnerable, subject to court enforcement or already insolvent, and errors land immediately. A seizure applied too broadly leaves someone unable to pay rent, one applied late exposes the bank, and a misidentified account holder suffers consequences they had no part in.

Automation raises throughput and, without measurement, raises the number of such cases proportionally. Nothing published describes error handling, how an affected customer is identified and made whole, or what safeguards apply where identity matching is uncertain.

The company's own framing names the institution's exposure, that backlogs in these processes expose banks to avoidable penalties. The value case therefore rests on absorbing a risk, with no stated undertaking about what happens when the platform itself causes one. No guarantee, indemnity or falsifiable commitment is published. For the customer whose account is seized, nothing at all is described. There is no notification, no correction route and no statement of who answers when an automated process reaches the wrong person.

Nothing about the underlying stack is disclosed. No model provider is named for the agentic components, no hosting arrangement or subprocessor list appears, and no external data source is identified despite these processes depending on court registers, enforcement authority filings and insolvency records.

The one relevant disclosure concerns people rather than technology, in a founding team drawn from investment infrastructure, consulting and applied machine learning research, which indicates where the process knowledge comes from.

The platform is described as replacing legacy back office tooling and as addressing fragmented systems and painstaking coordination across teams. That implies deep connection into core banking and case management, yet no system is named on either side. Nothing describes how cases arrive from courts or enforcement authorities, how account data is retrieved, or how outcomes are written back, and there is no public developer documentation.

Swiss banking secrecy and data location expectations are among the strictest in Europe, so the question arrives early. A platform processing customer enforcement and insolvency records for Swiss institutions will be asked where that processing occurs before any pilot begins. No hosting provider, region selection or residency commitment is published, and no private deployment option is described.

The outsourcing framing sharpens the pricing question. A platform sold as a service replacing headcount is naturally compared against the cost of that headcount or of an existing outsourcing contract. No pricing, packaging or basis of charge is published, and nothing indicates whether charge falls per case handled, per process, or as a managed service fee.

Buyers are banks and fintechs, and coverage at launch is deliberately narrow. Two processes are covered, account seizures and insolvency coordination, with chargebacks alongside. That focus is a sound strategy for a company this size and it is a real limit on what an institution can buy today. Geographic reach is unevidenced beyond a Swiss base and a European investor set. Because seizure and insolvency procedures are defined by national law, expansion into each new market means rebuilding the process rather than translating it.