◆ Autonomous Worker
Matches what the desk wants to move against which clients have shown interest, then builds and sends targeted axe runs per client. The matching, the message and the send run end-to-end behind a hard suitability and do-not-contact gate, with a compliance agent-as-judge clearing every send.
Memory
Working Current axe sheet matched to client interest signals.
Episodic Client holdings, prior interest, and response history.
Semantic Client mandates, sector preferences, do-not-contact rules.
Store CRM + axe history + client interest graph
Orchestration
router-fanout MCP
Harness · Managed Agents … scheduled + event-driven; client-comms tools.
Tools
{ } Inventory/axe feed API ⌕ Client CRM + interest signals Retrieval { } Client comms (chat/email) API { } Suitability + do-not-contact gate API
Evals & guardrails
- Outbound comms cleared by a compliance agent-as-judge before send.
- Suitability + do-not-contact rules enforced as a hard gate.
- Match relevance vs. resulting trades tracked per cohort.
Frontier edge
- ▲Proactive matching: predicts which client will want a line from holdings and flow signals, building the run before the desk asks.
- ▲Causal reasoning: models why a client traded last time (relative-value vs. mandate need) to target the axe that fits.
- ▲Self-improving fleet: shares which match patterns converted into trades across the sales-coverage agents between runs.
In numbers
900
Targeted runs / day
+2.4x
Axe-to-interest match lift
Handoffs
Fed by ← Credit RFQ Pricing Agent