Hazel

The removal test leaves nothing, because the product is interpretation rather than storage. It reads tax returns, pay stubs, account statements, emails and meeting notes, applies tax logic to them and produces a personalised plan in minutes, and separately answers questions across a firm's recorded conversations, documents and client systems alongside market and regulatory information. It captures and summarises meetings, drafts follow-ups, prioritises tasks and anticipates next steps. Every capability is model work; there is no underlying non-model application beneath it.

The current design keeps the adviser deciding. Output is described as adviser-ready reports and client-ready outputs, with the platform surfacing planning opportunities, drafting follow-ups and assembling analysis that a professional reviews before anything reaches a client, which is the correct construction where advice carries fiduciary weight.

The company has stated its next step explicitly, the ability to initiate actions directly on the custodial platform from within the assistant, which would move it from producing recommendations to executing them, and no threshold, approval or confirmation mechanism for that step has been described.

No accuracy, error rate, validation result or review statistic is published for a system that interprets tax returns and generates tax strategies, which is the measurement that matters most for this product. Deep tax logic is asserted without description of how it is maintained as rules change annually or how the platform is tested against known cases.

The disclaimer that no tax advice is provided means the vendor makes no representation about correctness, which is legally prudent and leaves the buyer without any published basis for assessing reliability.

This is the fastest adoption curve in the index. More than 1,000 wealth managers were using the platform within five months of its September 2025 launch, and following the February 2026 tax capability the company reported 1,600 net new and 2,000 total advisory firms subscribing within five weeks, with the founder noting that figure covered only the four weeks after the tax release and continued climbing. The company projects 1,500 further adviser sign-ups monthly. External validation is unusually hard: public markets repriced listed custodians on the launch, which is third party assessment no vendor can manufacture.

This is the second grade of its kind in the index and it goes further than the first. Two commitments are stated together: zero data retention arrangements with third-party artificial intelligence providers, and a strict commitment against using customer data to train.

The first is the one almost no vendor addresses, because it covers the path where client data leaves the vendor's own environment for a model provider's, and it names the specific contractual term negotiated rather than gesturing at security. A peer executive elsewhere in this index described exactly that negotiation as the hard part of deploying generative models on client financial data; here it appears as a published commitment.

The commitments are specific and multi-part rather than general: robust encryption, enterprise grade security, bank grade transport encryption standards, zero data retention arrangements with third-party model providers, and a stated refusal to use customer data for training. The platform runs on the same infrastructure as a self-clearing brokerage safeguarding billions in client assets.

Held at B because no data processing agreement, retention schedule, subprocessor register or residency position is published, and the holdings are unusually intimate, spanning tax returns, pay stubs and recorded client conversations.

A specific transport encryption standard is named rather than security being claimed generically, and the platform runs on the infrastructure of a self-clearing brokerage that safeguards billions of dollars in client assets, which is a materially stronger baseline than a software company's own environment.

Held at B because no attestation, certification, trust centre or penetration testing summary was located, and thousands of advisory firms now route client documents and recorded conversations through the platform.

No regulator, statute or rule is named for the platform. The parent is a regulated broker dealer and custodian, which is a different matter from the software's own position, and the published disclaimer states plainly that the firm and its affiliates do not provide tax advice and that investors should consult their own tax advisers. That is a sensible legal position and it leaves the regulatory status of automated tax planning delivered at this scale unaddressed.

Clients here are affluent, so access questions recede and a different exposure takes their place. Tax strategies are generated by interpreting documents and applying logic at scale across thousands of firms, and systematic error or a consistently aggressive interpretation would propagate identically across every client rather than varying by adviser judgement.

Complex or unusual circumstances, which are precisely the ones where tax planning matters most, are also where document interpretation is likeliest to fail. No analysis of accuracy across client situations, and no fairness or model behaviour disclosure, appears.

No guarantee, indemnity or correction process was located, and the published position is the opposite: the firm and its affiliates state they do not provide tax advice. Responsibility therefore rests entirely with the adviser who reviews and delivers the output, which is a defensible allocation and leaves the end client, whose tax return and pay stubs were interpreted by a model, with nothing described at all regarding notice, explanation or correction.

The company acknowledges third-party model providers exist in its stack and, more usefully, names the contractual term it has negotiated with them, zero data retention, which is more supply chain transparency than almost any vendor in this index offers. Held at B because no provider is identified, no base model named and no subprocessor register published, so a buyer knows the terms of the arrangement without knowing who the counterparty is or what would change if it were replaced.

Being inside the system of record is the deepest integration position available and this platform holds it. It draws on real-time account, beneficiary, household, holdings and balance data directly from a major custodian and unifies that with client relationship systems, emails, meeting notes and calendars, which the company states is a first for wealth management on both counts.

The strategic reasoning is stated correctly by its founder: for advisers the custodian is the ultimate system of record, where beneficiaries, cost basis and distribution data actually live, so accuracy depends on being wired into it rather than reading a copy. Additional custodian integrations are stated as next.

No hosting provider, region selection, residency commitment or private deployment option was located. The platform is described as running on the same secure infrastructure as its parent custodian, which conveys assurance by association without stating where processing occurs, and the third-party model provider relationships mean some processing necessarily happens elsewhere.

No price, tier or unit was located. Low cost is repeatedly identified by industry observers as a principal driver of the adoption surge, which makes the absence of a published figure more conspicuous rather than less, since the commercial proposition is explicitly that this capability is cheap enough to be adopted without deliberation.

Buyers are registered investment advisers and wealth management firms, and the platform is explicitly sold beyond its parent's own client base, stated as built for all advisers regardless of where they custody. Functional coverage spans meeting capture, client communication, research questions, task prioritisation, tax planning and firm-wide reporting on items such as required distributions and unrealised losses. The limits are that live custodial depth exists at one custodian today with others on the roadmap, and the market is United States advisory only.