Clearwater Analytics

The company argues that its AI is credible precisely because of what sits underneath it, a single validated data foundation rather than models bolted onto fragmented systems, and that argument is sound. It also settles this grade.

Clearwater's moat is the unified accounting, reconciliation and reporting backbone that carries more than ten trillion dollars in assets, and generative and agentic capability is layered across it, including embedded agents clients deploy for reconciliation, reporting and analysis and agentic tooling inside the risk platform. Apply the removal test and a complete institutional investment accounting and reporting platform remains, which is the business as it existed before any of this.

Control sits with the client in an important respect: agents are deployed by the institution into its own workflows rather than switched on by the vendor, and the company frames the goal as moving faster without sacrificing transparency, control or auditability, with the risk platform positioned to accelerate model validation that humans still perform. The counterweight is scale of delegation.

Clients are described as running hundreds of agents automating reconciliation, reporting, portfolio analysis and client communications, and no public material describes what review stands between an agent's output and an accounting entry, a regulatory report or a message sent to an investor.

Clearwater is on both sides of this axis, which is worth stating plainly. It sells model risk tooling: the risk and quantitative analytics platform embeds agentic capability specifically to accelerate model validation, exposure analysis, scenario work and tail risk modelling for institutional risk teams, and its cited rationale is intensifying regulatory scrutiny.

Auditability and transparency are named commitments, and a unified data foundation gives lineage that fragmented estates cannot. The gap is reflexive: a vendor selling validation acceleration publishes no validation evidence for its own models, no model documentation, no evaluation results and no stated position on supporting client review of its agents.

The evidence here is of a category no private vendor in this index can match, because Clearwater is publicly listed and therefore reports audited financials and operating metrics quarterly under securities law, where a materially overstated scale claim is a legal problem rather than a marketing one.

Platform scale is stated at more than ten trillion dollars in assets across insurers, asset managers, hedge funds, banks, corporations and governments, and named client deployments carry real detail, including a top ten German insurance asset manager with over fifty billion euros going live alongside a market data provider's buy side stack, and a credit focused manager implementing across a structured credit portfolio. The company also publishes annual primary research with a named external research partner. What is missing is per client outcome measurement.

The stewardship question here is unusually acute and follows directly from the architecture the company markets as its advantage. A single instance, multi tenant platform means competing insurers, asset managers and hedge funds sit on one system above a shared data foundation, and the AI is described as drawing its reliability from exactly that foundation.

Whether one client's holdings, trades or performance inform models or agents serving a competitor is therefore the central question a buyer must answer, and nothing public states the boundary, describes tenant isolation for model training, or says whether a client can decline participation.

The consumer privacy surface is structurally small, since the platform handles institutional portfolio, security, trade and accounting data rather than retail customer records, and that genuinely limits exposure under consumer financial privacy rules. The confidentiality stakes are elsewhere and are high: holdings, trading activity and performance for institutions competing directly with one another. This pass located no published privacy framework, data handling statement, retention schedule or subprocessor disclosure.

This pass located no trust centre, enumerated certification list, attestation scope or audit period on the public site. One verification route exists that is unavailable for the private vendors in this index: as a listed registrant the company is required to disclose its cybersecurity risk management, strategy and governance in periodic filings, and material incidents when they occur, so a buyer can read a mandated account rather than rely on a marketing page. That is a governance disclosure rather than a control attestation, and the grade reflects the absence of the latter.

Clearwater supplies technology and holds no financial licence, but it occupies an unusual position in the regulatory chain. Its software produces the investment accounting records, performance figures and regulatory reports that insurers and managers rely on for their own statutory filings, so an error propagates into a supervised submission rather than stopping at an internal report.

The platform is built explicitly against insurance statutory reporting, the European insurance capital regime and multiple accounting bases. The company is itself a listed registrant subject to securities disclosure and internal control requirements, which is a formal regime no private vendor in this index sits under.

Demographic fairness is not the live issue for a product whose subjects are securities and portfolios, so this axis reads here as accuracy governance, and the stakes are specific. Agents generate reconciliations, accounting output and regulatory reporting that feed statutory filings, and a systematic error would surface as a misstatement rather than an inconvenience.

This pass located no published accuracy figures for agent output, no evaluation methodology, no error or exception rate, and no description of the verification step between an agent's work and a filed number.

Listed status creates a form of accountability no private vendor here carries, since statements about the platform sit inside securities disclosure obligations and a material misstatement has legal consequence, and the company commits publicly to auditability and transparency. That is accountability for what it says rather than for what its agents produce. No accuracy guarantee, no remediation term, and no described process where agent generated reconciliations or reports feed a misstated regulatory filing.

Nothing public identifies which model providers power the generative and agentic layers, whether client portfolio and trade data reaches them, or which subprocessors are involved. The single instance multi tenant architecture makes that omission weigh more than usual, because competing institutions sit above one shared data foundation and the same undisclosed providers would serve all of them. Market data and pricing source dependencies are likewise not enumerated publicly.

The integration strategy is consolidation rather than connection, and at this scale that is the harder achievement. One system carries portfolio management, trading and execution, investment accounting, reconciliation, regulatory reporting, performance, compliance and risk analytics, replacing an estate a large institution would otherwise assemble from several vendors with reconciliation between them.

Where clients keep external components the platform interoperates, with a named live deployment running alongside a major market data provider's buy side stack, and acquired front office technology has been folded into the same platform for firms wanting order management inside it.

Delivery is cloud native on a single instance, multi tenant architecture, which is the source of the platform's consistency and also its residency problem. One instance serving European insurers subject to regional data protection and supervision expectations alongside North American and Asian institutions raises questions about where data physically sits and how transfers are handled that a per client deployment would not. This pass located no statement of hosting regions, residency options, transfer mechanisms or subprocessors.

No rate card, tier structure, billing unit or minimum is published and routes lead to contact paths. Listed status gives a buyer more than usual indirectly, since revenue, growth and retention metrics are disclosed in periodic filings and reveal the shape of the business, but none of that tells a prospective client what the platform will cost them. On the measure this axis applies, whether a buyer can size a deal without entering a sales process, the answer is no.

The broadest institutional coverage in this index. Clients span insurers, asset managers, asset owners, hedge funds, banks, corporations and governments, across public and private markets, and the functional span runs front to back from portfolio management and trading through accounting, reconciliation, performance, compliance and risk.

Regulatory coverage is correspondingly wide, addressing insurance statutory reporting and the European insurance capital regime alongside generally accepted accounting principles, international standards and tax bases. Operations run from the United States, United Kingdom, Germany and Hong Kong.