Streetbeat

The removal test leaves a data terminal. The models are patented, have run in production for three years, and are described as a multi agent architecture drawing on more than 170 real time data sets, integrating alternative data and modelling both action based and reaction based investing behaviour to anticipate short term market movements.

Agents perform the work rather than surfacing information for someone else to act on, automating investment execution, portfolio analysis, risk management and client engagement, and a firm can either take standard agents or have bespoke ones built. Nothing underneath survives their removal.

This sits at the high autonomy end of the wealth lane. The platform is described as agentic artificial intelligence for trading, with automated investing, auto trading options, model driven portfolio creation and dynamic rebalancing, which means agents executing rather than recommending. Control exists at design time, since a firm chooses standard agents or commissions ones built to its own requirements, and the lead investor characterises the effect as augmenting professional capability.

Nothing exists at decision time: no approval step before an agent trades, no threshold at which a decision routes to the adviser, no confidence exposure, and no description of what an adviser sees before a rebalance executes on a client's portfolio.

This vendor publishes something almost nobody else in this index does: a score against a named public agent benchmark, reporting 94.78 percent task completion on a suite that simulates realistic interactions where an agent must converse with a language model driven user, use interface tools against a database and follow complex policy guidelines.

That is external, reproducible evaluation rather than a self assessment, and adherence to policy guidelines under test is directly relevant to a regulated setting. Three years in production and patented models add track record. What holds it below the top grade is what the benchmark measures: it tests whether an agent can execute a task correctly and stay inside rules, not whether its investment judgement is sound, and no accuracy figure exists for the financial output itself.

Adoption is stated at 4,000 advisers across 15 countries, with recent expansion into Germany, Italy and South Korea, and one institution is named: a leading European online brokerage bank managing more than 120 billion dollars in assets, using the platform to strengthen its advisory offering.

Outcomes are given in the unit an advisory firm actually manages by, with advisers reported to have increased the number of clients they service fivefold and grown assets under management by up to 15 percent annually. The models have been in production three years.

A 15 million dollar Series A in late 2025 was led by the artificial intelligence fund of a state backed investment group, bringing total funding to 25 million, and its fund head describes the platform as combining a multi agent architecture with concrete adoption among advisers and institutions.

No data boundary statement was located. Agents operate across 4,000 advisers at many separate firms, several of which compete in the same markets, and the platform's value rises with observed outcomes, so whether portfolio decisions, client interactions or performance data from one firm inform the agents serving another is the material question. The option to deploy customised agents built for a specific firm implies some separation, and nothing states what it covers or whether the underlying models are trained in common.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. The platform integrates with customer relationship systems and, on the retail side, links directly to external brokerage accounts, so it holds client holdings, transaction activity and adviser interaction records across 15 jurisdictions with materially different data protection regimes. Nothing published describes how that material is retained or separated between the institutions sharing the platform.

No attestation, certification, trust centre or enumerated framework was located, with robust security protocols asserted without naming a control set or an independent assessment. A brokerage bank managing more than 120 billion dollars has completed vendor due diligence on this platform, so the assurance exists privately, and for a vendor whose agents execute trades on client accounts the published control set is what every subsequent institutional buyer will request first.

No supervisor, statute or instrument is named, and compliance appears only as an aspiration, in the stated ambition to be a leading compliant platform in the sector. The gap is wide given the footprint, because agents executing trades and constructing portfolios for advised clients across 15 countries engage investment services rules, suitability obligations and best execution requirements that differ materially between the United States, the European markets the company has just entered and its Asian expansion.

The inclusion case is the founder's stated mission, to make superior financial intelligence universally accessible and to democratise trading strategies previously exclusive to institutional investors, and the capacity figure gives it substance, because an adviser servicing five times as many clients can economically serve accounts that were previously too small to take. That is the threshold expansion pattern recorded across this lane.

Against it, the models are sold on predicting short term market movements and are also placed directly in retail hands with automated trading, and nothing published describes how portfolio construction handles clients whose circumstances fall outside common patterns, nor any analysis of outcomes across client segments.

No guarantee, indemnity or falsifiable commitment was located. The advising firm remains the regulated party and retains responsibility for what its clients receive, which is where accountability properly sits, and the benchmark result at least gives an institution some published basis for assessing agent reliability before deployment.

Nothing describes what happens when an agent trades or rebalances wrongly: no correction process, no notification path to the affected adviser or client, and no statement of what the vendor owes when its own execution is at fault.

The data layer is quantified but not identified, with more than 170 real time data sets stated as feeding the models and not one named, which matters because the alternative data sources behind a short term market prediction determine both its edge and its licensing exposure. The models themselves are the company's own and patented, which shortens the chain at the decisive point. No model provider is named for the conversational and agentic components, and no subprocessor list or hosting arrangement was located.

Deployment flexibility is the strength, with the conversational interface integrating into an institution's existing dashboards or running as a standalone terminal on desktop and mobile, so adoption does not require replacing what advisers already use. Customer relationship system integration is named as a capability, the retail product links to external brokerage accounts, and the platform ingests more than 170 real time data sets. What is not published is the list a wealth buyer needs, with no portfolio management, custody, planning or reporting platform named, and no developer documentation located.

No hosting provider, region selection, residency commitment or private deployment option was located. Operating across 15 countries including recent entries into two European markets and one Asian market means client portfolio and interaction data crosses several regimes with their own location requirements, and an institution deploying agents that trade on client accounts will be expected by its supervisor to know where that processing occurs.

No pricing, packaging or basis of charge is published. Two commercial structures are described, standard agents taken off the shelf against custom agents built to a firm's requirements, which almost certainly price differently and neither is quantified. The return case is framed instead through capacity and asset growth, which tells a buying committee what the platform is worth without indicating what it costs.

Four professional buyer types are served through one platform, spanning independent wealth managers, financial institutions, brokerages and financial applications, alongside a separate retail product. Geographic reach is real and evidenced at 15 countries with named recent entries into three distinct markets across Europe and Asia.

Functional coverage runs from portfolio construction and rebalancing through risk management and market analytics to client engagement and customer relationship integration. The deployment options widen it further, since the interface either embeds inside an institution's existing dashboards or runs standalone on desktop and mobile, so a large bank and a single adviser consume the same platform differently.