ebankIT

The digital banking platform runs web, mobile and branch channels without any model in the path. AI sits on top in three places: behavioral analytics and machine learning risk scores in the anti fraud layer, validation in digital onboarding, and a conversational assistant that carries out tasks for account holders. Remove them and the platform still opens accounts, moves money and runs campaigns.

The assistant acts only at the account holder's direction and inside each institution's operational rules. It can spot a redundant subscription or a suspicious transaction, then asks before freezing a card or taking any other step. On the fraud side, the risk thresholds at which the models escalate to an institution's team are unpublished, but an unusual login meets an extra verification step rather than a silent block, so a person stays in the loop at the point of friction.

An analyst working a fraud case gets the machine learning risk score and a graph view linking attackers, accounts and devices, enough to inspect one case. Behind that, no validation method, accuracy figure or false positive rate is published for the scores, and the assistant's answers carry no stated confidence or source, so model validation starts from the bank's own testing.

Hypothekarbank Lenzburg's redesigned channels, internet banking included, were built in three months on its Finstar core, the first partner bank after it went live in under two months, and updates ship every four weeks. Those figures measure delivery of the platform rather than results, and none of the named customers carries a figure tied to the fraud models or the assistant.

Coast Capital, a customer since 2018, moved hundreds of thousands of members onto the platform and has since launched a commercial service. Home Trust's executive vice president for digital and strategy reports higher engagement alongside lower costs, and Metropolitan Commercial Bank's digital onboarding won a Banking Tech Awards USA prize. Servus Credit Union, Alterna, Caisse Alliance and African Bank are named customers too.

The assistant runs on Azure OpenAI GPT-4.1 inside each institution's own rules and compliance guardrails, but nothing states whether conversation data is retained by the model provider or used for training. The behavioral fraud models learn from each person's interaction patterns, and whether that learning is pooled across institutions is not stated. That boundary matters more than usual for models built on how individual people type and navigate.

The privacy notice covers ebankIT's own relationships, from clients and suppliers to site visitors, staff and job candidates, under the EU General Data Protection Regulation (GDPR) and Portugal's Law 58/2019, with a data protection officer reachable by email. Platform data is handled under a GDPR privacy program with privacy risk assessments of workflows and anonymized test data in place of client data, described in general terms rather than as terms a bank could hold it to.

ebankIT's information security management system is certified to ISO/IEC 27001:2013, scoped to information security in the software development support cycle, and a SOC 2 report covers the security controls on development, releases and support of the platform. An independent firm runs penetration tests every year, and the company is appraised at Capability Maturity Model Integration (CMMI) Level 2. The certificate is cited against the 2013 edition, which the 2022 edition has since replaced, and the SOC 2 report type is not given.

ebankIT sells software and claims no financial license; the banks and credit unions it serves hold the authorizations. Its certification with Central 1 Credit Union covers the integration with Central 1's payment services, such as Interac e-Transfer and remote deposit capture, which is a technical certification rather than a regulatory status.

Behavioral analytics judge people by how they type, move a mouse and navigate, signals that vary with age, disability and familiarity with a device. Fairness testing for those models and for the assistant's recommendations is unpublished, and the governance material ebankIT does publish covers ethics, anti money laundering and anti corruption policies rather than its models.

Who bears the loss when the models miss a takeover or the assistant gives wrong guidance is not published. The checks that exist sit before the harm: the assistant asks the account holder before acting on a card or account, unusual logins meet an extra verification step rather than a silent block, and the platform is built to be audit ready.

The behavioral fraud models are presented as ebankIT's own modeling, and no subprocessor list is published. The assistant's stack is named in full: Microsoft Azure, with Azure OpenAI GPT-4.1 for language, Azure AI Speech for voice, and ebankIT's own Model Context Protocol connecting it to banking systems.

Public documentation for its application programming interfaces and marketplace listings are not published. The platform connects to existing cores through an integration layer built on those interfaces for traditional, hybrid and cloud native systems, and one core is named in production: at Hypothekarbank Lenzburg the layer connects to the Finstar core banking system, and the platform has since gone live at other Finstar partner banks.

Other named integrations include Central 1's payment services for Canadian credit unions, covering Interac e-Transfer and remote deposit capture, Alogent's deposit capture and item processing platform, and partners such as Entersekt, Prove and Unblu.

Hosting regions and residency terms are not stated for a platform that serves Canadian credit unions alongside European and African banks. The platform is cloud native, and the assistant runs on Microsoft Azure, which places the conversational workload with one named cloud.

Pricing comes only through a demo request. The platform is sold in modules under a recurring license model, so the shape of the purchase is visible even though no plan, rate or billing dimension is.

Retail banking, business banking and credit unions each have their own material, and retail, small business and commercial customers run on one shared stack. Institutions in 13 countries hold licenses, and Canadian credit unions get a dedicated layer connected to Central 1's payment services.