Nevis

Inference over unstructured material is the product: meeting summaries, pre meeting briefs and account opening submissions to custodians. Without the models, what remains is the fragmented stack the platform replaces.

Advice stays human led, and automation targets administrative work, yet within that scope client emails send automatically in the advisor's tone and account opening data is collected and submitted end to end. An approval step before an email sends, or a review gate on a custodian submission, is not described.

Consolidating a firm's records makes activity reconstructable, but that is data governance rather than model transparency. Summaries, drafted emails and account opening data all enter a regulated client file with no accuracy figure, evaluation method or error analysis published.

United Capital, GC Wealth and Apollon Wealth Management are quoted by name, and AlTi Tiedemann Global chose the platform in May 2026 for advisor workflows. RIAs on the platform manage more than $50 billion, yet no time saved or other outcome is tied to a named firm.

Customer data never trains or updates models, enterprise agreements are in place with OpenAI and Anthropic, and zero data retention terms bind the AI subprocessors. That retention term is written for Google Workspace email and calendar data, leaving meeting recordings outside its stated reach.

Processor terms reach the service: under the September 9, 2026 privacy policy the firm processes data for its advisory clients, retains it as each client directs and deletes it on the contract's terms, with a public subprocessor list. Retention periods themselves sit in each customer agreement.

Certification to SOC 2 Type II is stated, with encryption in transit, secure hosting and strict access controls described. The audit period, a report request path and any trust center are unpublished.

Fiduciary, recordkeeping and supervision duties over client communications sit with registered advisers, and the firm states no position on how automatically sent emails and meeting records meet them. Data protection under UK, EU and US law is what the privacy policy covers, not advisory regulation.

Emails written to match an advisor's exact tone go out under a named professional's identity, so a client cannot tell which messages the person wrote. Meeting summaries carry the usual speech recognition variance across accents, and nothing is published on client disclosure or on accuracy by speaker.

Audit records of user and attendee choices, acknowledgments and interactions let a disputed email or summary be traced. Neither an accuracy guarantee nor a remediation term covers a wrong summary or a miscollected account detail, and the client whose email was generated is given no published route to correct it.

The subprocessor list, updated July 30, 2026 with a change history, names Amazon Web Services as cloud and AI provider, OpenAI and Anthropic as AI providers and ElevenLabs for voice, each in the United States.

Unified search across a firm's systems and custodian submissions are claimed, and one customer cites client data spread over more than 30 systems. No CRM, custodian, reporting or planning system is named as an integration, and no developer documentation is published.

Personal data stays in its region of origin wherever feasible, US data in the US and UK and European data in the UK or European Economic Area, with standard contractual clauses for access from elsewhere. Its wherever feasible qualifier, and no private deployment option, push the detail into each contract.

A demo request stands where a rate, tier or billing unit would be. Pricing could follow advisors, assets or an enterprise agreement, and nothing indicates which.

Registered investment advisers from national firms with hundreds of advisors to multi family offices serving ultra high net worth families are the market, with a named customer of each kind. UK independent financial advisers also appear in the privacy policy, and nothing is built for institutions outside advisory.