EAGLES² — a modern, AI-powered successor to CAMELS

A Big 4 / top-3 consulting firm converted six months of fixed-scope work with Virtova into EAGLES² — a reusable, AI-native bank-rating platform that parallels and extends CAMELS across seven dimensions, 200+ specialized models, and 15+ external data feeds. The platform produces a defensible rating per institution without requiring internal-data access from the bank being rated, making it deployable at scale across the firm's hundreds of bank engagements.

Update: EAGLES² has since evolved into a live Frontier Foundry product focused on delivering daily regulatory-requirement updates against the seven-dimension framework. The public announcement is expected in mid-2026.

The firm runs a large financial-services advisory practice that supports hundreds of bank clients on safety-and-soundness, regulatory readiness, M&A diligence, and strategic-risk work. Every one of those engagements ultimately rests on a rating of the institution. The field-standard tool for that — the FFIEC’s CAMELS rating system — has well-known limitations.

The four problems with CAMELS in 2026

CAMELS is qualitative-heavy and examiner-dependent. Two examiners reviewing the same bank can produce materially different ratings. It is slow: a traditional CAMELS write-up runs to weeks of analyst time per institution. It under-weights the risk surfaces that dominate a modern bank’s risk register — cloud core banking, AI/ML model risk, quantum-era cryptography, fintech partnerships, and the cyber threats U.S. banks face today. And it is not reusable. Each engagement starts near zero, even when 80% of the underlying analysis repeats bank to bank.

The firm’s ask was specific. Give us a reusable, modern rating system we can deploy across hundreds of bank engagements without rebuilding the analytical stack every time.

A 1979 rating system cannot read the bank a 2026 board has to answer for.

EAGLES² — seven first-class dimensions

CAMELS measures six dimensions. EAGLES² measures seven. The framework was rebuilt around the risks that drive modern bank failure modes, with cyber and human capital rated as first-class dimensions rather than analyst commentary in an appendix.

The two S’s at the end are the framework’s signature. Cyber and human-capital risk are first-class rating dimensions, not footnotes. They are the two surfaces a 2026 bank board has to answer for both inside its own boardroom and at examination.

CAMELS → EAGLES² — what changes

EAGLES² is a parallel layer to CAMELS, not a replacement. It maps onto how U.S. banks are examined under the FFIEC framework, then extends the analysis into the seven dimensions a modern bank board has to manage.

How EAGLES² works — three layers

The platform delivers a defensible, evidence-backed rating per dimension and a roll-up institution score, generated from three layers working together.

Representative model coverage by dimension

The five external-data categories

The bank does not need to grant the firm access to internal systems for an initial EAGLES² rating to be produced. That property is what converts the engagement from a manual, one-off exercise into a productized capability that the firm can repeat across its full bank-client base.

The 6-month build plan

Fixed-scope, fixed-fee, on a tight timeline. Each phase shipped a concrete artifact the firm could deploy or hand off without external help.

Why the engagement succeeded

Regulator-grade methodology, built by regulators. The framework was designed by a team that has sat on the regulator side of the table — including former federal-bank-regulator leadership at Virtova — so the firm’s bank clients see a methodology that maps cleanly to how they are examined under the FFIEC framework today.

Reusable from day one. Every component — ingestion, models, external feeds, scoring — was built as a service the firm can re-point at the next bank in under a day, not the next quarter. Hundreds of clients become tractable, not aspirational.

No internal-data dependency. Because EAGLES² runs on regulatory and external data, the firm can produce an initial rating before access negotiations even start — accelerating sales cycles, diligence, and crisis-response engagements alike.

Quantitative defensibility. Cross-validation, credit-duration analysis, and multi-year back-testing against known supervisory outcomes give the firm’s partners audit-ready support for every conclusion.

Modern risk surfaces are graded, not narrated. Cyber, AI and tech modernization, and human-capital risk each carry a dimension score, not a paragraph in an appendix. The firm now opens board conversations on the risks the board itself carries, not the risks a 1979 framework happened to encode.

What the firm got at hand-off

What the firm can now drive across its bank-client base

• Faster engagements. Rating-first diligence in days. Analyst hours redirected to the advisory and remediation work the client pays the firm to deliver.
• Consistency across the practice. The same methodology applied to every bank, eliminating examiner-to-examiner drift on the firm’s own side of the table.
• New product offerings. Subscription rating products, quarterly bank-health briefings for boards, M&A target screens, and regulator-readiness assessments, all on a single engine.
• A point of view on modern risk. Cyber, AI, and human-capital scores the firm can put on the table in a board conversation from day one.
• Predictive lift over manual CAMELS. AI-augmented scoring catches deterioration earlier than analyst-only workflows, which gives the firm and its clients runway to act.

What this engagement is illustrative of

Dozens of consultancies will sell a Big 4 firm a slide deck on AI-enabled bank rating. A handful can build the analytical stack behind one. Almost none will ship a productized, reusable, regulator-grade rating platform — covering every U.S. depository institution, with 200+ trained models and 15+ external data feeds — inside a six-month, fixed-fee engagement.

The reasons that combination is rare are practical. A federated regulatory-data spine that survives examiner scrutiny has to be calibrated against historical supervisory outcomes from quarter one of its existence; that calibration is the work, and it is not optional. A 200-model library is a coordinated portfolio in which each dimension’s models work together under deterministic, explainable outputs, not a single model with versioned weights. And the discipline of building every input to be obtainable externally, so the platform deploys without internal-system access, is a constraint most teams design around rather than for.

The full intellectual-property stack travels. EAGLES², its seven-dimension rubric, the federated ingestion pipeline, the model library, the external-feed taxonomy, and the productization playbook are reusable across other Tier-1 advisory firms, foreign banking organizations supervised by the FRB, broker-dealers under SEC and FINRA scope, large credit unions, and de novo charters in their first examination cycles.

CAMELS is the framework U.S. examiners use, and it isn’t going away. EAGLES² is the parallel layer that runs alongside it, extends the analysis into the seven dimensions modern bank boards have to manage, and produces an evidence-backed score per dimension and a roll-up institution rating without internal-data access from the bank being rated. For a firm with hundreds of bank clients, that combination turns a recurring analytical exercise into a productized capability and gives the firm a defensible AI-grade methodology to bring into its next hundred bank conversations.

Continuing analysis

For continuing analysis on AI, audit, and U.S. banking regulation, the Sultan Meghji Substack publishes weekly to over 13,000 subscribers. Related insights on the regulatory frame this engagement was scoped against: Cyber audit in the neural-net era, AI governance for U.S. banks after the FDIC years, and NIST AI RMF in practice.