Timeline

An award winning planning package and a discretionary investment business would both run fully without the models. Everything sold before May 2026 survives intact: the withdrawal rate and cashflow engine that was the whole product for eight years, the factfind, the digital letters of authority and the inheritance tax planner. So do the model portfolios and multi asset funds carrying over three billion pounds, the portfolio analytics, and the platform handling order execution and custody.

The assistant is one named item inside the planning menu, launched in May 2026. The vendor's own framing is AI embedded in an integrated ecosystem rather than bolted onto a disconnected stack. That is fair as a description of how it is wired, but it describes where the models sit rather than how much of the product they carry.

The boundary is published, categorical and covers every write path. The assistant is stated never to send, submit, change or trigger anything independently. Every output is reviewed by the adviser before it goes anywhere, and every action requires explicit approval. Naming four verb classes in one sentence makes the statement of the limit complete on paper. It is asserted as a design principle, though, and nothing describes the mechanism behind it.

Nothing published describes an administrator control over the gate, a record of what was generated against what was approved, or an audit trail a compliance reviewer could examine. Nor is there any way for a firm to verify the boundary holds rather than take it on trust. A stated principle and an enforced configurable control are different things to a firm that has to evidence its supervision.

No accuracy figure, error rate, benchmark or validation method is published. There is also no grounding or citation guarantee tying a generated statement back to its source, and no drift statement or revalidation cadence. The unmeasured risk is concrete because of what the assistant produces.

Structuring a financial plan from uploaded documents and generating a client report that reflects both a meeting discussion and portfolio analytics means assembling figures that end up in a suitability file. A plan drafted from a misread document is wrong in a way the reviewing adviser is poorly placed to catch, since the draft looks complete and the source document is no longer in front of them.

The adoption evidence is substantial and it is entirely about the investment business rather than the software or the assistant. Assets under management passed one billion pounds within roughly two years of launch and three billion pounds by 2025. That is money actually placed by advice firms, a harder signal than a testimonial. The company also runs its own adviser conference and publishes an awards page.

For the software and the assistant, nothing is published: no named advice firm reporting an outcome from the planning software, no measured time saving, no adoption figure for the assistant and no customer account of using it. The assistant launched in May 2026, and the vendor states advisers are already using it without naming one.

The same organization holds the client records, the portfolio positions and the platform custody data, so material flowing through the assistant comes from a wider pool than a standalone note taking tool would ever touch. Nothing published describes the boundaries around that pool.

The assistant announcement, product pages and launch coverage say nothing, in either direction, about whether recorded meetings, transcripts, generated drafts or client records are retained, used to improve the service, or used to train, tune or evaluate any model.

What the assistant does raises the questions a privacy policy would need to answer. It joins client meetings to record and summarize them, extracts indicators of client vulnerability, and draws on client records, portfolio holdings and platform data in the same operation. Vulnerability indicators about an identifiable individual are among the more sensitive categories a firm holds. The product and announcement material carry no retention, deletion or processing terms. A privacy policy is linked from the site footer, and its terms, like any processing terms offered to advice firms, are not reflected here.

Product pages, the assistant announcement and the site navigation name no attestation, information security standard, audit report or trust page, and no security page is linked. The gap matters more for Timeline than for a pure software supplier. The group operates a custody platform and holds client assets, which brings operational resilience and outsourcing obligations to the authorized entity, and the assistant now processes recorded client meetings. Those obligations sit with the regulated company rather than the software company, and nothing published bridges the two. The regulated entity's own published disclosures are not reflected here.

Timeline draws the regulatory line itself, in its own footer. Investing and platform services are provided by one company, named with its registration number and its authorization and reference number at the Financial Conduct Authority. Planning software and tools are provided by a second, separately numbered company, and the published statement that this entity is unregulated is set in bold. The assistant sits inside the planning entity.

So a real authorization exists within the group, and it covers discretionary investment management and custody rather than the generative tooling. Supervision of one part of a group does not read across to software sold by another, and Timeline says so plainly rather than leaving a reader to assume the authorization covers everything.

The assistant automatically extracts indicators of client vulnerability from meeting conversations, and that capability goes undiscussed. Under United Kingdom consumer protection expectations this is a determination with direct consequences: a client correctly identified receives adapted service, and a client the system fails to identify does not.

Detection of that kind depends on how people express difficulty, which varies with age, health, language, culture and how willing someone is to disclose. Uneven performance across client groups therefore produces uneven protection for exactly the people the requirement exists to protect. No evaluation of vulnerability detection across client types is published, nor any account of what the system treats as an indicator, nor any fairness testing.

The oversight design carries the allocation of responsibility implicitly. A boundary under which the assistant never acts without explicit adviser approval is a real control, and it also places every consequence of a wrong draft on the person who approved it. The vendor states plainly that the firm remains in control at all times. No warranty, accuracy commitment, correction obligation or explicit allocation between vendor and advice firm is published.

The client has no route at all: no described means to learn that a system recorded a vulnerability indicator about them, to see it, or to contest it, and that indicator then shapes the service they receive.

No model provider, model family, version, inference host or subprocessor list appears on the assistant page, in the launch announcement or its trade coverage, or in the product navigation. The vendor's positioning makes the silence conspicuous. Its argument against competing tools is that they are generic and disconnected while this one is connected to real client data, an argument entirely about where the data comes from and never about where the reasoning happens. A reader is told precisely which internal systems feed the assistant and nothing about whose model consumes them.

The connectedness is real and it points inward. The assistant's stated advantage is that it reaches across planning, platform, portfolios, reports, workflows and client records without the adviser moving between systems. That depth exists because the vendor owns every one of those surfaces. It is also the limit: an advice firm running a different back office gets nothing from it, because the connections run to this vendor's own stack rather than to the estate a firm already has.

Outward reach exists elsewhere in the business, through model portfolio distribution across fourteen platforms, digital letters of authority that connect to providers, and an account aggregation integration. No programmatic interface or named connection into third party CRM systems is published.

The products are reached through a browser with a hosted authentication provider, and nothing about topology or location appears in the product material. There is no public statement of hosting arrangements, a storage region or a residency commitment, nor of where recorded client meetings and generated content are processed. Timeline sells exclusively into United Kingdom advice firms handling United Kingdom client data, so a residency statement is the disclosure a buyer would expect to find first.

Timeline publishes no rate, tier or per adviser figure, and no statement of whether the assistant carries its own line or is included in a planning subscription. No fund or portfolio charge appears in the product material, and separate fund charge documents are not reflected here. The pricing question matters more because the company sells software, discretionary portfolios and a custody platform to the same buyer.

A firm cannot establish from published material whether software pricing moves depending on whether that firm also uses the portfolios, the arrangement most worth understanding before adopting any part of it.

The breadth comes from product range rather than buyer variety. The buyer is a United Kingdom financial advice firm and nothing else: one jurisdiction and one professional buyer, with no bank, insurer or asset manager segment addressed. The international footprint the company described in earlier years is not visible in current positioning.

To that single buyer type the company sells planning software, discretionary model portfolios, a multi asset fund range, portfolio analytics and its own custody platform. An advice firm can take one component or the entire stack. The portfolios also reach firms that use none of the software, through distribution across fourteen third party platforms.