Flybits

Models decide what each customer sees and when, which is the product's purpose, and the 2025 agentic capability goes further by assembling journeys across cards, loans, deposits and insurance around a life event rather than executing a campaign. The research arm is developing predictive journeys and content intelligence.

Held at B because the underlying platform is content and offer delivery infrastructure with experience templates and channel integration, which would function on rules rather than models, more crudely but intact, so this is a delivery platform whose targeting became model driven.

The agentic capability is described without any accompanying limit. Agents guide customers through major life events and unify cards, loans, deposits and insurance into single journeys, which means a system is assembling multi product financial recommendations for a customer at a consequential moment, and no human review, approval workflow, suitability check or escalation route appears anywhere in published material. Nothing states what a bank's product governance sees of which customer received which recommendation, or on what basis.

Measurement is offered as a platform capability, with institutions able to design, launch and measure experiences and a 2025 release improving performance visibility, and none of the resulting figures are published. No engagement lift, conversion rate, recommendation accuracy or validation result appears, and nothing describes how the agentic system's product suggestions are tested for appropriateness before reaching customers.

The investor list is the strongest strategic backing in this index, comprising a global card network, the venture arms of two major banks and established technology funds, meaning three financial institutions assessed and funded the company directly.

It has reached a late stage round, operates from three cities across two continents, ships quarterly product releases including new market and language support, launched an agentic capability in 2025, and runs a research arm partnering with academic institutions. Held at B because no bank or credit union customer is named anywhere, and no adoption, engagement or conversion figure is published, so investor conviction substitutes for deployment evidence.

No boundary statement was located. The platform sits between institutions and their customers across banking, cards and payments, observing what content is delivered and how customers respond, and a library of pre-packaged experiences implies patterns learned somewhere are reused elsewhere. Nothing states whether engagement outcomes at one institution inform recommendation models serving another, or what a bank contributes about its customers by adopting the platform.

Privacy is presented as a design property rather than a compliance afterthought, with the company stating that experiences are delivered while preserving customer privacy, and it names zero party data as a methodology, meaning information a customer deliberately and knowingly provides rather than behaviour inferred without their involvement. That distinction is meaningful for a personalisation platform, since the usual model depends on inference.

Its own consent handling offers granular opt out. Held at B because no data processing agreement, retention schedule or subprocessor list was located, and nothing states how zero party data is balanced against the behavioural signals that context awareness necessarily requires.

No attestation, certification, trust centre or enumerated framework was located. Major banks and a card network have conducted diligence as investors, and any deploying institution would have completed a supplier assessment covering a platform embedded in its customer facing application, so the underlying controls have been examined and none of the documentation is published.

No regulator, statute or rule is named, with regulated industries referenced only as a market descriptor. That is a notable gap for a platform that recommends credit, deposit and insurance products to retail customers across North America and Europe, since product promotion, suitability, fair treatment and marketing consent are governed differently in each of those markets and the agentic capability makes the recommendation more consequential rather than less.

The dual use question is the same one that arises for any system timing offers to personal circumstances, and here it is sharpened by the company's own language. Journeys are organised around major life events, and the research arm describes exploring predictive habit formation journeys, which means the platform is designed to reach people at moments of change and to shape recurring behaviour. Those are the moments when someone is most persuadable and least deliberate. Nothing published describes safeguards against offering credit at a point of vulnerability, no fairness testing appears, and no analysis of which customers receive which offers is disclosed.

No guarantee, indemnity or correction process was located. The customer receiving personalised offers has nothing described: no statement that they are told why a particular product appeared, no route to correct the inferences driving it, and no stated ability to opt out of behavioural targeting while continuing to use their bank's application. With agents now assembling multi product journeys around life events, the absence of any disclosure or objection mechanism is more consequential than for conventional marketing.

No base model, provider, hosting arrangement or subprocessor is identified for the agentic capability or the personalisation engine. Inputs are the institution's own customer data supplemented by contextual signals, which keeps the dependency inside the customer relationship, and the company does not describe what those contextual signals are or where they come from, which matters because context awareness usually implies external sources such as location or third party data.

One named digital banking platform integration exists, which matters because reaching customers requires sitting inside the application they already use rather than alongside it. The stated methodology is connect once and deploy widely, with delivery across all channels and explicit emphasis on minimal technology involvement so that digital teams rather than engineering can ship experiences, which is the practical constraint at institutions whose development capacity is committed elsewhere. No core banking, card processing or customer data platform is named beyond that single integration.

No hosting provider, region selection, residency commitment or private deployment option was located. The company operates across North America and Europe and added new market support during 2025, which brings jurisdictions with differing requirements on where customer behavioural data may be processed, and published material does not address any of it.

No pricing, packaging or basis of charge was located. The commercial argument is explicitly about cost, promising dramatically reduced ongoing development costs through a connect once approach and lower product acquisition costs through digital channel delivery, with no figures attached and no indication whether charge follows customers reached, experiences deployed or institution size.

Buyers span banks, credit unions, card issuers and payments businesses, with the company describing itself as serving regulated industries generally, and coverage extends across the full cardholder lifecycle from activation through engagement, spend and retention as well as retail banking experiences. Geographic reach spans North America and Europe from three offices, with new markets and languages added during 2025. Held at B because no institution is named in any market and the depth of deployment in each is not described.