Crealogix

A digital engagement platform trading since 1996 with one named model product among three named solutions, the other two being a funding portal and a lending origination hub. Strip the models and the entire business remains: the modular hub, the mobile applications, the open banking interfaces and the two lending products.

Included on the Clearwater precedent because the conversational product operates inside a regulated servicing channel, handling customer communications with encrypted messaging and a retained audit trail rather than sitting outside the bank's record. The disclosure is thin: what the product actually does with models, beyond combining artificial intelligence with automation for personalised communication, is not described anywhere located.

No gate, threshold, confidence measure, escalation path, handover rule or sampling audit is described. The product is characterised by combining models with automation for real time customer interaction, which is a statement of what it does rather than of how it is controlled. For a conversational system speaking directly to bank customers, the central oversight question is when it stops and passes to a person, and nothing located answers it.

No accuracy, precision or recall figure, benchmark or validation method was located, and none of the four properties this index accepts as evidence of model risk discipline is present. Nothing describes intent recognition accuracy, misrouting rates or how often the system answers a customer question wrongly.

The audit trail records what was said, which supports reconstructing an individual exchange and says nothing about how often exchanges go wrong across a corpus, the same per item explicability recorded against PerformLine.

A named client, a United Kingdom wealth manager that selected the platform to build its client portal and mobile experience for high net worth clients, plus the counted institution base restated by an acquirer paying a disclosed price. Off an A because no quantified outcome attaches to any named client and nothing measures what the model product achieved: no deployment count for the conversational solution, no containment or resolution rate, no effect on servicing cost. Industry awards were not credited, on the same basis as the submission based recognitions declined elsewhere in this index.

No statement was located on training data, retention of conversation content, or whether models learn across the institutions on the platform. The gap is specific for this product type: a conversational service handling customer communications for more than 600 institutions accumulates the actual language customers use about their money, which is exactly the corpus that would improve a shared model and exactly the corpus a bank would want ringfenced. The audit trail is presented as a compliance feature and nothing describes who else may read what it retains.

A published data protection policy exists and cites the legal bases it relies on under European data protection law, names its use of processors and joint controller arrangements, and points to its security certifications. That is more than most of this pocket publishes and it concerns the company's own corporate and website processing rather than the product. Nothing describes what customer conversation content the platform retains on behalf of a bank, for how long, or what the bank's own customers can see or request.

Named frameworks with the correct nouns and an independently verifiable entry: the company states certification under ISO 27001 alongside SOC 1 and SOC 2 reporting, and a group entity appears in a certification body register under the 2022 revision of the information security standard with a dated certificate and a stated scope covering cloud services, managed services, network and security and data centre operations.

Off an A because the register entry covers one group entity rather than the platform as a whole, the SOC reports are referenced rather than shown to be available to prospects, and no trust centre, penetration testing summary or subprocessor disclosure was located.

A software vendor holding no financial licence of its own, with no supervisory programme, sandbox admission or regulator run assessment of the model product located. Its clients carry the authorisations. Open banking interface support means the platform operates inside regimes its customers are regulated under, which is product scope rather than a credential held by the vendor.

No framework, bias testing, fairness evaluation, model documentation or independent assessment of an artificial intelligence management system was located. A conversational system serving retail, private, corporate and small business customers across 15 countries handles many languages and registers, and comprehension quality varies systematically across them. That is the fairness question for this product shape and nothing published addresses it.

Nothing published describes liability, indemnity or recourse when the conversational product tells a bank customer something wrong. The exposure is direct rather than mediated by an adviser: this system speaks to the end customer, so an incorrect answer about a balance, a payment or a product term reaches the person immediately with no professional between them and the model. Nothing describes whether the customer is told they are speaking to a machine, or how a disputed exchange is escalated.

No base model, provider, version or hosting arrangement is named for the conversational product, and no statement was located on whether third party foundation models sit underneath. The product name announces that models are involved and the public material stops there. An institution recording which system produced a customer facing answer, which is basic practice for a regulated servicing channel, cannot do so from anything published.

Genuine integration positioning with a large deployed base: a modular platform designed to sit on the interaction layer above whatever core a bank already runs, unifying internal and third party systems through open interfaces, with open banking support and the ability to add partner components alongside the vendor's own, plus a stated alliance with a core banking provider.

Held at B rather than A because the specific systems coordinated are described generically rather than enumerated, and the depth of connectivity into custody, payments and risk is not evidenced the way it is for the front to back and engagement platforms graded A in this pocket.

Described as a modular cloud platform delivered as a service, with no region choice, hosting location, tenancy model, on premises path, residency commitment or subprocessor list located. One certified group entity has a scope covering data centre infrastructure and operations, which indicates the company runs hosting itself without disclosing where. For a vendor serving Swiss and Middle Eastern banks under localisation expectations, that is a material silence.

No pricing, band or rate is published for the platform or any of the three named solutions. Modular packaging is disclosed in structure, so a buyer can see the units of purchase, and nothing indicates what any unit costs or how the conversational product is licensed relative to the hub it runs on.

More than 600 customer institutions across 15 countries, a figure restated by the acquirer in transaction materials rather than appearing only in vendor marketing. Segments are named individually and span retail banking, private banking, corporate banking, small and medium business banking, wealth and investment management and the public sector, across Europe, the Middle East and Asia Pacific. Breadth is anchored to counted institutions and counted countries, which is the discriminator that separated PerformLine from Saifr.