HyperNorm AI

The removal test leaves an adviser reading research alone, which is the stated problem. The architecture is disclosed in unusual technical detail for a company this young, combining language model agent systems with symbolic artificial intelligence and causal reasoning, and that choice is functional rather than decorative: causal modelling supports risk propagation analysis showing how a market event moves through a specific portfolio, and the symbolic component is what allows the reasoning behind a recommendation to be set out rather than inferred.

The adviser decides throughout and the product is positioned as a co-pilot rather than a decision maker. Output is prioritised alerts flagging which portfolios may need attention, an explanation of the factors behind each alert, and suggested actions with rationale aligned to the client's mandate, all of which the adviser reviews before acting.

Mandate alignment is itself a suitability control, since a recommendation is generated against the client's stated investment constraints rather than in the abstract. Held at B because no threshold, override mechanism or escalation rule is described, and scenario simulation and recommendation generation run without any stated checkpoint.

Explainability here is an architectural commitment rather than a feature claim, which is the distinction this axis rewards. The company states plainly that rather than simply generating outputs the platform explains why a recommendation is made and how specific market developments affect the portfolio, and the combination of symbolic reasoning with language models is precisely the design that makes that reasoning inspectable rather than reconstructed after the fact. Causal modelling over correlation is a further deliberate choice. Held at B because no accuracy, backtest, validation result or evaluation method is published for any of it.

The company reports paying clients across the United States, Singapore and India, which is genuine commercial traction on three continents for a business founded in 2024, and a 2.2 million dollar seed round co-led by the deeptech arm of an established technology group and a specialist AI fund.

Beyond that the record is thin: no client is named, no count of firms or assets is given, no outcome figure of any kind is published, and the product is described as having moved from beta with initial revenues flowing.

No boundary statement was located. The platform observes portfolio positioning, mandates and adviser decisions across firms competing for the same clients, and nothing states whether that material informs models serving others, whether client data is isolated per firm, or what an advisory firm contributes by adopting the platform.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. The platform holds client portfolio holdings, investment mandates and financial goals, and processes them across three jurisdictions with materially different data protection regimes, none of which is addressed. An independent analysis of the company identifies this specifically as a principal risk to its growth.

No attestation, certification, trust centre or enumerated framework was located. Advisory firms in three markets have engaged commercially, so some assurance was given privately, and nothing is published for a prospective firm to assess before connecting client portfolio data.

No regulator, statute or rule is named. The buyers are registered investment advisers and wealth managers operating under suitability, fiduciary and disclosure obligations in three separate jurisdictions, and a platform generating portfolio recommendations sits directly inside that regulated activity, so the absence of any mapping is a gap the buyer's compliance function would need to close itself.

Clients are affluent, so access questions recede and model behaviour takes their place. Recommendations generated at scale from shared models mean that if the reasoning favours particular instruments, sectors or strategies, that preference propagates identically across every adviser using the platform rather than varying with individual judgement. Nothing describes how recommendation patterns are monitored, whether outputs are tested for systematic tilts, or how a firm would detect that its advice had converged with everyone else's.

No guarantee, indemnity or correction process was located. The advisory firm retains fiduciary responsibility for any recommendation it acts on, which makes an explicit statement of what the vendor stands behind more useful rather than less. The end client, whose portfolio is monitored and whose recommended actions originate from the platform, is not addressed at any point.

The architecture is described by type, naming language model agent systems, symbolic components and models trained on millions of data points, which is more structural candour than most offer, and no base model, provider, hosting arrangement or subprocessor is identified. Market and macro data sources feeding the signals are likewise unnamed, and for a platform whose output is portfolio recommendations the provenance of those inputs determines both coverage and reliability.

No named integration, interface documentation or connected system was located. A platform monitoring live multi-asset portfolios and mapping market signals onto holdings must connect to portfolio management or custodial data to function, and how it does so, and with what, is not described anywhere.

No hosting provider, region selection, residency commitment or private deployment option was located. The company operates from India with paying clients in the United States and Singapore, which makes processing location an early procurement question for advisory firms holding client data under their own jurisdictions' rules.

No pricing, packaging or basis of charge was located. For a platform sold to advisory firms of varying size, whether charge follows advisers, portfolios monitored or assets covered determines which firms can adopt it, and none of it is published.

Three distinct buyer roles are served with different needs met by the same engine: advisers managing client relationships, portfolio managers, and chief investment officers, with the platform reaching registered investment advisers, wealth managers and fund managers. Asset coverage spans equities, fixed income, structured products and other classes, and paying clients exist in three markets across North America and Asia. Held at B because volumes are unstated in each market and no regulatory adaptation between them is described.