Explainable AI Next-Best-Action recommendations for retail banking sales and customer care.
Several signals compete for the same morning: maturities, balance drops, overdue items, card expiries, callbacks.
Due dates don't compare urgency, value, service risk and recent contact.
The action must fit the evidence, current policy and what already happened.
Data abundance creates decision scarcity.
For this project BIDV is treated as not yet having an AI Next-Best-Action capability. AI features shown are proposals.
Steps 1–2 and 7–8 are solid system-of-record capabilities, and they stay. We redesign only the decision stage.
No shared decision layer applies common ranking and suppression; context isn't packaged around the action; feedback is logged as an outcome, not as a quality signal.
Turn available data into an explainable, governed, feedback-driven decision product inside the RM workflow. Keep human judgement for the decision itself instead of spending it on assembling information.
Retail Relationship Manager · 31 · 6 years in retail banking
“I need to know which few customers matter today.”
“Can I trust the reason? Has someone already called? Is this policy still valid?”
Checks tasks, profile, events and history before deciding. Reopens policy documents for the important cases.
Pressured and wary of black-box suggestions. Reassured by evidence, fresh data and the option to say no.
explain a recommendation in under 30 seconds using facts the RM can verify?
Rules handle hard eligibility, permission and suppression. AI is used where ranking and combining patterns adds value.
Rules run before ranking. Every decision is logged with its reason code, evidence timestamp and model/rule version.
Profile, events, products, activity
Persona/need, NBO, churn — hypotheses
3–4 approved action types for the pilot
Permission, eligibility, recency, expiry
Urgency, relevance, outcomes, workload
Why-now from traceable facts
Create next activity only after the RM confirms — drafts are never auto-sent
Review action, reason, outcome, versions
Adoption, rejection, drift; pause by type
Illustrative, not to scale. RM snooze and “not relevant” feedback feed back into monitoring and tuning.
BIDV-NBA · fictional placeholder data
Every 2-week cycle: review rejection reasons, false positives, stale-data cases, interviews, prep time and incidents, then re-prioritise by evidence.
Relationship managers
Premier customers
Maturity · card expiry · retention · overdue support
Then scale, redesign, narrow or stop, per action type
without quality incidents
a healthy rise, not a quota
improves, or holds while saving time
not relevant, already contacted, wrong timing, missing context
the primary safety constraint
Trust is tracked with a short recurring survey and interviews, never inferred from acceptance alone.
One ranked queue with evidence instead of several lists
Guardrails + ranking + explicit human review
RM judgement becomes visible and learnable
Signal, freshness, policy and outcome around each pair
Monitoring by action type; pause what doesn't work
No ROI claim yet: a cost-aware plan and a baseline, not a made-up payback period.
“BIDV-NBA doesn't replace the Relationship Manager. It helps the RM focus their attention, understand why, and choose the next action with better context.”