FNZ

Custody, trading, advice and reporting have run since 2003 without models, and the first model product only shipped in August 2025 as the opening of a planned series. It matters because its output, meeting preparation and conversation prompts, shapes what an adviser puts to a client in a regulated interaction.

Advice AI is presented as help for advisers, who still interact with the client and give the advice, and FNZ plans Microsoft 365 Copilot and intelligent agents for its own middle and back office work. No gate, confidence display or sampling review is described for generated material before it reaches a client. The client facing advice path is where a misstatement would do harm.

Advice AI and the analytics built on FNZ's data warehouses come with no published accuracy figure, benchmark or validation method. The Analytical Data Warehouse mixes traditional data science, machine learning and generative AI, so a client's model risk team would need the method behind each service it adopts before relying on the output.

Aviva built and launched a new customer experience on FNZ in a nine month deployment, and Santander moved a legacy wealth proposition to a direct to consumer service tied into its retail bank. Vanguard and Colonial First State are named with their programs as well. Nine months measures how long the platform took to deliver rather than anything the AI did, and no named client yet reports what Advice AI or the productivity suite achieved.

FNZ answers the pooling question directly: its Aggregated Data Warehouse combines anonymized data from all clients, and that pooled set underpins its AI productivity suite. How the data is anonymized, how long it is kept and whether an institution can keep its clients out of the pool go undescribed, so the pooling is disclosed without the terms that would govern it.

FNZ publishes no privacy program, retention position or data subject rights framework for platform data. That matters because holdings, transactions and personal circumstances for nearly 30 million investors sit on the platform, most of whom deal only with their own institution, and the Aggregated Data Warehouse combines anonymized data from every client.

Client assets for nearly 30 million people sit with FNZ, yet it publishes no security certification, attestation report or trust center. Any credential surfaces only inside procurement, which leaves end investors with nothing to check.

FNZ sits inside its clients' regulatory perimeter as a regulated firm in its own right. FNZ (UK) Ltd and FNZ TA Services Limited are authorized by the Financial Conduct Authority, its Irish firm FNZ (Europe) by the Central Bank of Ireland since December 2020, and FNZ Custodians (Australia) carries its own disclosures. Those authorizations cover custody, administration and investment services rather than the AI layer, which no regulator has tested.

A large multi market study on AI in wealth management and calls for responsible deployment describe how the industry should govern models. How this vendor governs its own is undisclosed, with no framework, bias testing or model documentation.

A model on Microsoft's AI platform, inside FNZ's administration service, used by an adviser at an institution, reaches an end investor four parties removed. Who is liable when generated material misstates a client's position or prompts an unsuitable recommendation goes undescribed, and those four parties make it the first question for the contract.

The Microsoft partnership names FNZ's stack: Azure AI Foundry at the center of the platform, Azure OpenAI models through Foundry, Microsoft Fabric for data and analytics, and Microsoft 365 Copilot agents for FNZ's own operations. An institution still cannot record which model produced a given client communication, because the model, version and region behind Advice AI go unnamed.

The platform is the system of record end to end, from custody and trading to advice and reporting, rather than a layer above one. Modular wealth components are distributed through a hyperscaler partner's marketplace, extending reach outward.

Local data centers, client service and operations run in each region, and Microsoft Azure is FNZ's cloud partner for AI. Which regions hold which clients' data, how tenancy is separated and where the AI models run are unpublished, across dozens of countries with differing outsourcing rules.

Platform contracts are negotiated per client over multi year terms, typically on assets and transaction volume, with no published rate. Whether the model product is bundled into platform economics or charged separately is unstated, and disclosed group revenue speaks to scale, not cost.

More than 650 financial institution partners and nearly 30 million people use the platform, with over US$2.5 trillion in assets on it. Banks, wealth managers, insurers and asset managers are all served, and each region gets local data centers, client service, operations and compliance.