Agents / Customer Analytics & CRM
Lead Follow-up Agent
Turns scored leads into follow-up tasks for relationship managers. Each morning it takes new and re-scored leads from lead-priority-scorer, removes customers without marketing consent or on the do-not-call list, balances the load across each RM's pipeline, and drafts the next action and talking points for every Priority A and B lead in the CRM. Part of the optimised lead tracking use case.
Ownership and sensitivity
Who approves access
- Owner, Customer Analytics & CRM, Mei Ling Tan
Business-sensitive. Models and data products scoped to named business units.
Entitlement per business unit, approved by the owner; conditions attach.
- Customer PII
- Contains or processes personal data about customers (PDPA 2010).
- GenAI
- Generates text; prompt screening for customer data and output review apply.
Tools
- Lead scoring (lead-priority-scorer)
- BigQuery read (crm-leads-history)
- BigQuery read (customer-360-mart)
- Consent and do-not-call registry
- Gemini on Vertex AI (policy-rag-copilot)
- CRM: task writer
Data access scope
- Read-only: open leads, lead scores and the customer's product holdings
- Read-only: consent flags and the do-not-call list
- Write: follow-up tasks and draft talking points in the CRM for the assigned RM only
Guardrails
- Leads without marketing consent or on the do-not-call list are never assigned a contact task
- No more than 12 new tasks per RM per day
- Talking points never mention protected attributes or the raw model score
- Halts if the lead score feed is older than 24 hours
Human-in-the-loop
- RM decides whether and how to contact each lead
- Branch manager can reassign or drop tasks before 10:00 each day
- Weekly review of conversion by priority band with Customer Analytics
Orchestration
Every tool call is scoped by the declared data access above; the orchestrator cannot reach systems outside it. Owning business unit: Consumer Banking.
Live demo run
Watch the agent execute a real scenario step by step, every tool call, validation, and human checkpoint is traced and auditable. Typical run: ~9.1s.
Audit
Updated 2026-07-23. Run traces retained 24 months for audit under the platform governance policy.
Community · 0 threads
Questions, findings and requests from the business units that use this asset. Owners reply here; threads with upvotes surface to the owning team's inbox.