RSK Engine room · 6 desks · 14 agents

Risk Management

Measures exposure, sets the limits, makes the call within the board's appetite.

Credit, market, liquidity, operational and model risk: the second line that measures exposure, enforces limits, validates models and runs the regulatory stress tests across the whole book.

How it runs

Always-on monitors watch every limit and exposure in real time, aggregate counterparty risk across the whole book, set and enforce the limits, decide the marginal credits, and draft the validation and stress narratives … all within the appetite the board sets. Oversight agents gate the high-stakes calls before commit; the control plane audits every decision after.

Explore Risk in the city →

Credit Risk

3 agents

Measures and manages the risk that borrowers and counterparties default … from single-name reviews to portfolio concentration and rating maintenance.

Workflow · Spreading financials → rating / scoring → annual & event-driven loan review → covenant monitoring → watchlist → portfolio concentration analysis.

Loan Review Agent

Runs periodic and event-driven credit reviews on the lending book.

Spreads the financials, re-rates the borrower against the bank's scorecard, and writes the credit review memo citing the line items that moved the rating. Runs the annual-review queue continuously and flags deteriorating credits as the financials shift.

Covenant & Early-Warning Agent

Continuously watches the book for covenant breaches and deterioration signals.

Watches the signals that change credit risk … covenant tests, payment behaviour, adverse media, equity and CDS moves … and raises a watchlist nomination the moment a borrower starts to deteriorate. Runs as a live early-warning loop across the portfolio.

Counterparty Exposure Agent

Aggregates counterparty credit exposure across the whole trading book.

Pulls every derivative, repo and SFT position against a counterparty across every system, nets it under the master agreements, and emits a single current and potential-future-exposure number. Recomputes live on each position change, not on an end-of-day batch.

Market Risk

2 agents

Measures the risk that market moves hit the trading book … VaR, sensitivities (Greeks), stressed VaR, and trading-limit enforcement.

Workflow · P&L + risk capture → VaR / sensitivity calculation → limit checks → breach investigation → P&L attribution & back-testing.

Limit Breach Monitor

Watches every trading and risk limit in real time and triages breaches.

Sits across the desks and watches VaR, sensitivity, notional and concentration limits as they move intraday. On a breach it pulls the contributing positions, attributes the cause, and hands the desk-head agent a breach packet with the driver already identified.

VaR & Sensitivity Surveillance Agent

Surveils VaR, Greeks and back-testing for anomalies and explains the moves.

Watches the daily risk numbers for anomalous moves … a VaR jump with no obvious cause, a back-testing exception, a sensitivity that drifted from the prior day … and writes the attribution. Runs the P&L-attribution pass overnight and surfaces the commentary by open.

Liquidity & Treasury Risk

2 agents

Measures funding and liquidity risk … LCR/NSFR, intraday liquidity, the funding profile and the cashflow ladder under stress.

Workflow · Daily LCR/NSFR computation → cashflow-ladder build → intraday liquidity monitoring → funding-concentration analysis → regulatory reporting.

Liquidity Ratio Agent

Computes and surveils LCR, NSFR and the funding profile daily.

Aggregates the HQLA stock and net cash outflows across every entity and currency, computes LCR and NSFR with full lineage, and attributes the drivers when the ratio moves. Emits a continuously-reviewed liquidity pack with the drivers attached.

Intraday Liquidity Monitor

Watches intraday cash and collateral positions across payment systems in real time.

Tracks the bank's cash position across every payment and settlement system through the day, projects the cashflow ladder forward, and warns of a liquidity pinch or a stuck payment before it stalls the queue. Reprojects on each flow change and alerts the treasury funding agent.

Operational Risk

3 agents

Manages the risk of loss from failed processes, people, systems and external events … RCSA, loss-event capture, key risk indicators and scenario analysis.

Workflow · Risk-and-control self-assessment (RCSA) → loss-event capture & categorisation → KRI monitoring → issue & action tracking → scenario analysis.

RCSA Facilitation Agent

Drives the risk-and-control self-assessment cycle across the business.

Pre-populates each unit's RCSA from prior assessments, incidents and control-test results, drafts the risk and control descriptions, and challenges assessments that the evidence does not support. Produces an evidence-backed draft; the risk-oversight agent owns the ratings.

Loss-Event Capture Agent

Detects, captures and categorises operational-loss events automatically.

Watches incidents, GL adjustments, customer complaints and reconciliation breaks for the signature of an operational loss, captures it, maps it to the Basel event category, and links it to the failed control. Books the event into the loss database directly from the source feeds.

KRI & Issue Tracking Agent

Maintains key risk indicators and chases open issues and actions to closure.

Refreshes every KRI from source, flags the ones drifting toward their threshold, and chases owners on overdue remediation actions. Escalates the stuck actions, not the whole register.

Model Risk Management

2 agents

Independent oversight of the bank's models … inventory, validation, ongoing monitoring and documentation under SR 11-7. Includes validating the AI agents elsewhere in this very building.

Workflow · Model inventory → validation (conceptual soundness, outcomes analysis, benchmarking) → ongoing performance monitoring → documentation → annual revalidation.

Model Validation Agent

Runs independent model validation analyses and drafts the validation report.

Reproduces the model developer's results, runs the benchmarking and outcomes analysis, probes the assumptions, and drafts the validation report to SR 11-7 structure. Spawns parallel test sub-agents for replication, benchmarking and sensitivity; an independent validation oversight agent owns the challenge and the sign-off.

Model Performance Monitor

Continuously monitors live models for drift and performance degradation.

Watches every model in production … including the agents in this building … for drift, stability breaks and performance decay, and raises a revalidation trigger when a model degrades past tolerance. Runs as an always-on monitor per model and routes triggers to the validation agent.

Stress Testing (CCAR/DFAST)

2 agents

Runs the bank's regulatory and internal stress tests … scenario data prep, loss/PPNR projection orchestration, results aggregation and the supervisory narrative.

Workflow · Scenario receipt → data preparation across portfolios → model-run orchestration → results aggregation → narrative & regulatory-pack drafting.

Scenario Data-Prep Agent

Prepares and reconciles the data inputs for every stress-test run.

Assembles the portfolio, balance-sheet and macro-variable inputs each stress model needs, reconciles them across systems, and documents every assumption and mapping with full lineage. Gates the release of each input set on a data-quality oversight agent before the model run proceeds.

Stress Narrative & Reporting Agent

Aggregates stress results and drafts the supervisory narrative and reg pack.

Aggregates the loss and PPNR projections into capital outcomes, attributes what drives the results under each scenario, and drafts the supervisory narrative and FR Y-14 schedules. Drafts but never submits; an oversight agent gates the filing within the board's mandate.