RightCapital

The calculation engine carries the product, and the vendor's own architectural statement is the clearest evidence for it. An engine covering retirement projection, tax aware distribution modeling and the rest was the entire product from 2015 and remains so, with both inference capabilities arriving in 2026. The company states that every agent output originates from that engine rather than from external sources.

That is an admirable control, and it also describes where the substance lives: the engine computes and the agent surfaces. Remove the models and advisers build plans exactly as they did through 2025, with more typing. Document extraction in the import capability does carry real throughput, and it feeds a platform that was complete without it rather than one that goes empty in its absence.

Two real controls are published and a third is absent. The first is scope. The agent reviews, surfaces and simulates rather than sending, filing or transacting, so the exposure is a wrong conclusion rather than a wrong action, and that is a design choice rather than a limitation. The second is governance of access, since a firm decides at organizational level which advisers and assistants may use the agent at all, which is a supervision lever.

What is missing is any review step between an agent surfaced anomaly or proposed strategy and its arrival in a plan a client will see. Nothing published shows how an adviser's own assumption is told apart from one the agent proposed, and no audit trail of what the agent did is described.

A published architectural constraint addresses the exact failure mode a planning agent has. The company states that all agent outputs originate from its own calculation engine, so the agent surfaces figures the verified engine computed rather than producing them itself.

Applied to a product whose answers are retirement probabilities and tax outcomes, that is a structural defense against a fabricated number reaching a client plan, and it is specific and checkable rather than a policy assertion. The constraint covers where figures come from, and nothing else is measured. There is no extraction accuracy rate for the import capability and no evaluation of whether the anomaly detection finds what it should or flags what it should not. No benchmark, validation method or revalidation cadence is published either.

An independent research citation is present, and it is a placement rather than a measurement. The company cites a named productivity study by a widely followed independent research house for the claim that it holds the highest user satisfaction among advisers, and names the study and its year, which is more disciplined than the usual unsourced superlative. Against that, the citation dates from 2024 and reports a ranking without a published figure.

Independent survey data from 2026 covering nearly three thousand advisory firms places two competing planning platforms ahead on market share, with this vendor's own share unstated. The one quantified operational figure, a reduction in manual data entry of at least 70 percent, is self reported with no named firm behind it.

There is no statement either way on whether client profiles, uploaded documents or generated plan content are retained, used to improve the service, or used to train, tune or evaluate any model. One published statement invites the wrong conclusion. The company says insights come exclusively from its own verified engine rather than external sources, which describes where answers come from and says nothing about where client data goes. A reader scanning quickly could take it as an assurance that nothing leaves the platform, and it is not that.

How uploaded client documents are handled once their contents have been translated into plan inputs is undescribed, and that question is specific to what shipped this year. A document read by the import capability has served its purpose the moment its figures land in the plan. Whether the source file persists afterwards matters to a firm, and nothing states it. No retention schedule, deletion commitment or processing terms are reflected here either, though a privacy policy may cover them.

This profile carries no named attestation or information security standard, and no audit report or trust page. For a platform holding complete household financial profiles for thousands of advice firms, and now ingesting uploaded client documents through its import capability, a published position almost certainly exists.

A software supplier holds no license and requires none. The output sits close to a regulated act without performing one: a retirement projection and a recommended strategy shape the advice a client receives, and the recommendation is made by the adviser under the firm's own permissions. Nothing published says what a firm retains to evidence how a plan was constructed, or how a reviewer would tell an assumption entered by an adviser from one proposed by the agent. The second question is the live one, because the agent proposes strategies by varying assumptions and nothing describes how that provenance is recorded.

Nothing published evaluates fairness, tests output across client circumstances or explains how the agent decides what to raise. The exposure sits in selection rather than in any explicit judgment about a person. A review that surfaces anomalies, questionable assumptions and future gaps is choosing which of a household's many imperfections deserve an adviser's attention, and a builder that returns three strategies from a probability target is choosing which three out of a very large space. A consistent tendency in either would steer whole categories of client toward particular retirement ages, spending levels or product choices without any adviser noticing a pattern across their book.

The vendor publishes no warranty, accuracy commitment or correction obligation, and responsibility between vendor and advice firm is not allocated. The grounding constraint removes one exposure and leaves the others untouched: an agent that cannot invent a figure can still miss one. The check that scans a profile for missing information is the clearest case.

A gap it fails to flag becomes a plan built on incomplete data, with an adviser reasonably believing it was checked, and nothing published addresses what the firm may rely on that check to have covered. The client whose plan was reviewed has no described route to see which conclusions came from the agent.

No model provider, model family, version or inference host is named for the agent, the document extraction or the assistant that answers platform questions. The claim that outputs originate from the proprietary engine rather than external sources reads at a glance like a declaration that no third party model is involved. It is a statement about the provenance of answers, not the provenance of the reasoning. Nothing published establishes whose model interprets a document or composes an explanation.

Interoperability is evidenced in three different forms and enumerated in none. A named two way relationship exists with a leading meeting assistant vendor. That is the connection that matters most in this market, because it closes the loop from a client conversation to a plan input. A migration tool exists for importing from a specific competing planning platform, which is a harder engineering commitment than an ordinary integration.

And the import capability ingests arbitrary uploaded documents, which is interoperability with the world rather than with a system. An integration directory naming the client relationship systems, custodians and account aggregators an advice firm runs alongside this is likely published and is not reflected here.

The platform is delivered through a browser, with a client portal alongside the adviser interface. Hosting, storage region, tenancy and residency go undescribed, as does where household financial profiles and uploaded documents are processed and stored. The vendor may publish some of this.

The packaging is disclosed and no number is. The planning agent carries no incremental charge and is instead gated to the upper two subscription tiers, so the commercial question becomes which tier a firm needs rather than what an AI add on costs. No tier rate, per seat figure or comparison of what each tier includes is reflected here. The company is generally understood to publish a pricing page, and its figures are not included in this profile.

A large single profession base in a single jurisdiction, addressed across tiers rather than across institution types. The buyer is a financial advice firm, thousands of them, served through a subscription ladder whose upper tiers carry the agent, so solo practitioners and larger firms are both addressed with different entitlements rather than different products. Two limits matter.

Everything is bound to one country's tax and retirement rules, and no bank, insurer or asset manager is addressed as a buyer in its own right. The migration tool aimed at a competitor's user base is a distribution mechanism rather than a coverage extension.