Addepar
Addepar is a New York headquartered data and portfolio intelligence platform for investment professionals, founded in 2009 and serving more than 1,400 firms across nearly 60 countries who manage and advise on close to 9 trillion dollars of assets, of which over 40 percent sits in alternatives. Buyers span registered investment advisers, multi family offices, single family offices, private banks, institutions, asset owners, allocators and fund managers, with named clients including Morgan Stanley, AllianceBernstein and the Geneva multi family office Lobnek Wealth Management.
The platform unifies multi custodial data, illiquid alternatives and complex legal entity structures into a single portfolio view, and layers analysis, reporting, financial planning, projection through Navigator, and a trading capability launched in late 2024 that replaced the rebalancing technology acquired with AdvisorPeak. The open platform integrates with roughly 650 software, data and consulting partners. Offices run through New York, Salt Lake City, London, Edinburgh, Geneva, Dubai, Pune and Sao Paulo, with about 250 staff across Europe and the Middle East, against annual research spending stated at over 150 million dollars.
The artificial intelligence line has three shipped parts. Addison, launched in March 2026 and available to firms that opt in, is a native experience embedded in the platform that answers natural language questions about performance drivers, exposures and liquidity, understands portfolio structure, alternatives complexity and firm specific entitlements, and returns outputs the company describes as permission aware and traceable, tied directly to the underlying records. Alts Data Management applies machine learning with human verification to extract and normalise private markets data.
Intelligent Statements converts unstructured financial statements into review ready files. The company states that Addison originated in research aimed at building a proprietary large language model trained on its own accumulated historical data, and it acquired the workflow automation startup Arcus in May 2025 and folded that capability in. Announced next steps are proactive insight delivery with market aware context and cited sources, plus agentic workflows across data operations, analytics, forecasting and reporting.
Capability Axes
Capability grades
15 of 15 axes rated · 5 graded A or B
The Clearwater shape, cut the same way as the host vendor of this pocket. The company now calls itself a data and artificial intelligence platform, but the asset underneath is fifteen years of multi custodial aggregation, entity modelling, reconciliation, reporting and trading, and stripping the models leaves all of that intact and still the reason firms buy.
The Oxane refinement is worth applying here because it complicates the picture rather than settling it: the public markets side keeps working without inference, while the alternatives side degrades, since private markets data reaches the platform through machine extraction of unstructured statements. Mixed, and C is the honest placement for the whole.
Three real controls, consistently stated across the vendor's own material rather than in one press release. Access is permission aware and entitlement aware, so the assistant sees only what the querying user may see, and firm specific entitlements are honoured. Outputs are traceable and tied directly to the underlying records, which means an adviser can follow any answer back to the source rather than accepting it.
And the capability is opt in at firm level, so it is off until a firm turns it on, which is a genuine deployment control and rarer than it sounds. Held off A because the shipped product analyses rather than acts, so no gate has yet had to be described; the announced agentic workflows are said to keep humans in the loop and that claim names no gate, no threshold and no escalation path; and unlike the reference case on this axis there is no confidence surfacing beside the provenance.
One property carries this and it is a real one rather than a value statement: every response is stated to be traceable and tied directly to the underlying records, so an output can be followed back to the portfolio data that produced it. That is checkable in a demonstration and it is the explainability route this index credits, distinct from a vendor asserting explainability as a noun.
Held off A because nothing else is published: no accuracy or error rate, no validation approach, no benchmark, no monitoring regime and no drift policy, for a system answering questions about performance, exposure and liquidity that advisers will repeat to clients. The announced citation of sources is a roadmap item and is not credited.
Clears the bar squarely, which is uncommon. A named client case study, a Geneva based multi family office, reports roughly an 80 percent reduction in manual workload after moving onto the platform, so a named customer and a quantified outcome are joined rather than published in separate places.
Beyond that the installed base is named at the top end, with a global wire house and a major asset manager among the clients, and the scale figures are specific and consistent across sources at 1,400 firms, nearly 60 countries and close to 9 trillion dollars. Recorded honestly: the case study is published by the vendor, and the quantified figure is the customer's reported experience rather than an independently audited measurement.
A new and pointed shape for this axis, and it runs opposite to every other instance in the index. Almost every vendor here either says nothing about training or asserts that client data is never used for it. This one states affirmatively that its assistant originated in research aimed at building a proprietary large language model trained on its own accumulated historical data, and then publishes nothing further.
The word carrying the weight is securely, which describes a method and answers none of the questions that follow: what that accumulated data comprises, whether it is client derived, whether the firms whose portfolios constitute it agreed to that use, whether anything is isolated between clients, and whether a firm can decline while still using the platform.
Opt in is published, but it governs use of the assistant rather than contribution to what trained it, and conflating the two would be an error. Call it the affirmative training claim with no terms: a disclosure that is more candid than the norm and leaves a buyer worse informed than a plain denial would.
A permissioned data model with granular role based access control is published, and encryption in transit and at rest is stated, which is where the disclosure stops. No privacy framework compliance statement, no processing addendum, no sub processor list and no retention or deletion terms were located.
Notable for a platform holding the complete balance sheets, entity structures and beneficial ownership of ultra high net worth households across nearly 60 jurisdictions, which is among the more sensitive collections in this index.
Nothing is credited, and the reason is a formulation worth adding to the credential presentation catalogue as a new negative shape. The vendor's own page states that the platform aligns with industry frameworks such as the second service organisation control standard at type 2. That sentence borrows the authority of a named standard without claiming to hold anything: aligns with is not holds, and such as makes even the list illustrative rather than complete.
A buyer learns neither whether a report exists, nor its type, scope or period. Real controls are stated alongside it, encryption in transit and at rest and a permissioned model with granular role based access, and those are recorded without moving the grade.
Banked check, cheap and near certain to move this: a platform carrying a global wire house as a client has passed that firm's vendor review and will hold current reports; look for a trust or security page and a procurement documentation portal.
No licence, no supervised test of the product and no programme enrolment. The trading capability brought in house in late 2024 is the one place where registered status might attach and nothing published addresses it, so it is recorded as an open question rather than credited. Banked check.
Nothing published on fairness testing, performance variation or governance of outputs. Consumer protection exposure is genuinely low, since this is portfolio analytics and reporting for professionals managing the assets of wealthy households rather than any decision about an individual's access to a financial service.
The residual concern is narrower and worth stating: an assistant that surfaces what matters most across a book of business is prioritising which clients and which holdings get attention, and in a wealth business attention is the service being sold.
No published allocation of responsibility for a wrong output and no route for any affected party, who in this case is the underlying household rather than the paying firm. The exposure is indirect but not trivial: an assistant answering questions on exposure and liquidity feeds analysis that reaches a client, and the traceability recorded on the model risk row helps an adviser check an answer without giving anyone a remedy when the answer was wrong and acted upon.
The standing pattern holds that building permits silence, and this is a build. The company states the model is proprietary and trained on its own data, which tells a buyer the provenance is internal rather than a third party interface call, and that is fractionally more than the usual bare claim of proprietary technology.
Below that top line there is nothing: no base model, no architecture, no version, no hosting arrangement and no supplier, for a system the vendor describes as a large language model, which in practice is very rarely built from nothing.
Depth by function and by counterparty. The platform is the aggregation layer itself, pulling multi custodial feeds, illiquid alternatives and complex legal entity structures into one reconciled view, which is the operational core for the firms that run on it rather than an adjacent tool.
On top of that sits an open integration estate stated at roughly 650 software, data and consulting partners, with specific named integrations published, including a financial planning tool that builds plans from live Addepar holdings. A trading capability was brought in house in late 2024. The combination of being the system of record and publishing an integration estate of that size is what separates an A from a B on this axis.
Nothing published on hosting, region, tenancy or residency. The gap is more consequential here than for a domestic vendor, because this platform holds portfolio and beneficial ownership data for firms in nearly 60 countries, including European and Swiss private banking clients operating under bank secrecy and data protection regimes where residency is a routine procurement question. Eight offices across four continents is a corporate footprint, not a data residency statement, and the two should not be confused.
No price, tier, unit or pricing basis published anywhere. An independent research profile characterises the platform as more expensive than its competitors, which is third party commentary and earns nothing under the source test but is worth recording as the only pricing signal that exists in public.
Among the broadest in the index and matching the host vendor of this pocket. More than 1,400 firms in nearly 60 countries with close to 9 trillion dollars managed or advised on the platform. Buyer types are genuinely distinct rather than restatements of one another: registered investment advisers, multi family offices, single family offices, private banks, institutions, asset owners, allocators and fund managers, ranging from boutiques to a global wire house. Asset coverage spans public and private markets with over 40 percent of platform assets in alternatives, which is the hard half. Eight offices across four continents.
Alternatives to Addepar
The closest documented capability profiles to Addepar in the same categories, ordered by similarity across the same fifteen axes the index grades every vendor on. Closest documented profile, not a claim that either product does the same job. No vendor pays for placement.
Documents AI Centrality where Addepar does not
Documents AI Centrality where Addepar does not
Documents Deployment Model and Data Residency where Addepar does not
Documents AI Governance and Bias Disclosure where Addepar does not
Documents Model Supply Chain Disclosure where Addepar does not
A lighter documented profile than Addepar
Similarity is computed axis by axis from published grades, not from a composite score. The index does not aggregate grades into a total. See the fifteen axes and the methodology.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
No pricing data has been verified for this vendor. Pricing information will be published here once confirmed through vendor disclosure or third-party estimation.