Lifecycle Campaign Manager AI

Invoked when: any of the 25 monitored borrower events fires in the LOS or CRM

• –Monitors the LOS and CRM event stream for the 25 borrower events that signal a lifecycle moment requiring a campaign response: disbursement (welcome journey), first EMI completion (activation confirmation), third EMI (early loyalty acknowledgment), sixth EMI (top-up eligibility milestone), twelfth EMI (annual relationship review and offer), EMI bounce (immediate soft-touch retention message), DPD entry (collections handoff coordination with Early Bucket Caller AI), restructuring agreement (empathy and support sequence), loan closure (renewal and next-product journey), and referral submission (reward confirmation and encouragement). Each event fires a pre-approved journey template in under 15 minutes.
• –Manages journey state per borrower — each borrower is enrolled in at most one active journey at any time. When a new event fires for a borrower already in a journey, the agent evaluates whether the new event should interrupt the current journey (an EMI bounce interrupts a promotional journey immediately), continue the existing journey (a repayment milestone during a welcome sequence adds a milestone step without restarting), or queue the new journey for post-current-journey enrolment. Journey collision logic prevents borrowers from receiving contradictory communications simultaneously — a borrower in a retention journey does not also receive a promotional campaign.
• –Personalises journey content at the message level: the borrower's first name, the specific loan product they hold, the EMI amount and due date where relevant, and the tenure remaining. Personalisation is drawn from the CRM and LOS at message dispatch time rather than at journey design time — this ensures that amounts and dates in messages are always current even if the journey was designed weeks earlier. Stale data in campaign messages is a significant trust and compliance risk; the agent re-queries live data at dispatch, not at enrolment.

Invoked when: daily churn scoring cycle runs — all borrowers within 90 days of loan closure or with identified churn signals are scored

• –Scores each active borrower on churn probability using a composite of behavioural signals: declining app login frequency, reduced WhatsApp message response rates, EMI payment timing drift (paying later in the cycle each month without bouncing — a leading indicator of financial stress), absence of engagement with prior campaign communications, and proximity to loan closure without a renewal inquiry. Churn prediction 60 days out gives the retention journey enough time to intervene before the borrower has already decided to go elsewhere for their next loan.
• –Classifies borrowers by churn risk tier — High (churn probability above 70%, immediate retention journey enrolment), Medium (probability 40–70%, softer engagement sequence with relationship value messaging), and Low (probability below 40%, standard lifecycle journey continues without modification). High-tier borrowers receive a personalised outreach from the relationship manager within 24 hours in addition to the automated campaign sequence, because high-value borrowers at churn risk warrant human contact, not just digital automation.
• –Tracks churn prediction accuracy at the cohort level — the percentage of High-risk borrowers who actually lapsed in the subsequent 60 days, compared to the Low-risk baseline. Where prediction accuracy falls below 65% for a cohort, the churn model inputs for that segment are flagged for review by the Customer Insights Agent AI and the marketing head. A churn model that cannot predict accurately is worse than no model, because it misdirects retention spend onto the wrong borrowers.

Invoked when: a borrower is onboarded to a product segment, or when the marketing head updates a segment journey template

• –Maintains separate journey libraries for each product segment: MSME borrowers receive journeys calibrated to business seasonality (GST filing periods, working capital cycles, festive production runs) rather than the personal finance calendar; retail personal loan borrowers receive journeys anchored to individual repayment milestones and personal financial planning moments; gold loan borrowers receive shorter, higher-frequency journeys reflecting the typically shorter tenure and different renewal dynamic of secured gold products; co-lending portfolio borrowers receive journeys that reflect the bank-NBFC co-lending context, with messaging that aligns to the co-lender's brand standards where required.
• –Updates segment journeys when the product or credit policy changes: if the gold loan LTV limit changes, all pending gold loan renewal journeys that reference LTV are updated to reflect the current parameter before dispatch. Journey templates that contain product-specific parameters are linked to the policy engine so that stale parameters in active journeys are caught and corrected automatically when the underlying policy changes — rather than discovered after incorrect information has been sent to borrowers.
• –Reports segment journey performance monthly: open rates, click rates, journey completion rates, and the conversion outcomes attributable to lifecycle campaigns (top-up acceptances, renewal applications, referral submissions). Attribution is based on a 14-day attribution window from campaign exposure to conversion event — conversions that occur outside the window are not attributed to the campaign, even if temporally proximate. This conservative attribution approach ensures that campaign ROI reporting reflects genuine campaign-driven conversions rather than coincidental repayment cycle activity.