Capital markets in one cockpit — live, cited, explainable.
Hedge teams stitch P&L, lock outcomes, ratesheet drift, and pricing across spreadsheets, OMS exports, and email. RateStack's Hedge Cockpit collapses that into one real-time view: P&L roll-ups, daily sparklines, pullthrough scoring, an AI daily briefing that cites what moved, and the Exception Inbox routing the anomalies to triage.
The Hedge Cockpit materializes a lock-outcome view from the same NATS event stream that drives pricing and locks. P&L, sparklines, exception severity, and pullthrough predictions all read from the same source-of-truth — no reconciliation overhead, no parallel ETL. Every number drills into the underlying event via correlationId.
The AI Desk Assistant sits on top of the cockpit. Ask 'why is the 7yr cohort drifting' and you get a cited answer with links into the audit chain. Suggested resolutions for exceptions land as drafts in the Inbox; operators accept, edit, or reject — never the model deciding on its own.
Shareable URL state means a desk handoff is a link. The morning meeting's briefing renders as a PDF straight from the page. Compliance can ride along on the same URL with read-only capability scoping; nothing gets re-exported.
Before vs. after
The shape of a day.
The same operating model, rebuilt around explicit pricing and a single audit log.
Before
Morning P&L roll-up stitched from OMS exports, lock spreadsheets, and the LO platform.
After
Cockpit aggregates server-side; SSE pushes updates. Hover any number to drill into the source event.
Before
Pullthrough is a Monday-morning model the desk reruns manually.
After
Pullthrough scored on every lock with explainable attribution; cockpit shows trend live.
Before
Exceptions live in email and a Slack channel; nothing resolves with a paper trail.
After
Exception Inbox streams via SSE with AI-suggested resolutions and audit-chained resolve / dismiss / embed / recompute.
Before
Daily briefing is a Word doc someone reads at 7am.
After
AI daily briefing renders in the cockpit with citations into source events; PDF export hits the inbox before standup.
Capabilities, framed for you
The platform pieces you'll touch first.
Hedge cockpit
P&L, pullthrough, anomalies, AI briefing — one server-rendered page.
Exception Inbox
SSE-pushed exceptions with AI suggestions and audit-chained workflow.
AI desk assistant
Conversational, cited, audit-chained AI on top of your desk's data.
Event-driven primary feed
NATS JetStream backbone — same stream feeds cockpit + webhooks.
Sell-side pricing
Per-investor completion audit for hedging and bid-tape reconciliation.
Historical replay
Reprice any prior moment to validate hedge inputs deterministically.
Onboarding
What week one looks like.
A pragmatic sequence — from sandbox to first signed quote.
- 1
Day 1: connect the events
Subscribe to pricing.computed, locks.*, and exception.* webhooks; mirror them into your hedge stack via the correlationId.
- 2
Day 2: light up the cockpit
Cockpit reads from the materialized lock-outcome view; first session shows live P&L the moment the materializer warms up.
- 3
Week 1: pullthrough validation
Run the scorer against your last 90 days; reconcile against your funded outcomes. Tune any custom features.
- 4
Week 2: AI briefing tune
Adjust the daily briefing to match your morning meeting's vocabulary. Citations stay grounded.
- 5
Week 3: Inbox routing
Route exception severities into your alerting (email, Slack on Business+). Custom detectors register on Enterprise.
- 6
Production
Cockpit becomes the desk's primary morning surface. Reconcile pullthrough monthly; the audit chain replays any decision.
The cockpit is the first time the entire desk reads from the same numbers in real time. The AI briefing actually saves us 20 minutes every morning.
Frequently asked
Specific to your operating model.
Do we have to migrate off our OMS?
No. The cockpit and the events run alongside whatever OMS you have. Webhooks deliver the same data into your existing hedge stack with HMAC signing and DLQ replay.
How is the AI grounded?
Strictly on your tenant's lock, pricing, exception, and ratesheet data. Citations are server-verified — an unresolved citation degrades the reply with an audit warning.
Can multiple desks share the platform?
Yes. Desks are first-class entities; capability scoping gates cross-desk visibility. URL state encodes the desk + filter set.
What's the cost model for AI?
Rate-limited at 30 rpm per tenant by default; monthly cost ceiling on every tier. Exceeding the ceiling routes to a degraded mode (template-based) rather than billing surprise.
Can we add custom exception detectors?
On Enterprise, yes. Custom detectors are JVM functions registered against the event stream; they run alongside built-ins. Talk to sales for the SDK pattern.
Ready when you are
See hedge desks on RateStack.
Live demo with your real ratesheets, your real scenarios, and an honest read on whether the platform fits your team.