Pre-Approval Offer AI

Invoked when: weekly portfolio scan cycle runs, or repayment milestone triggers real-time eligibility check

• –Scans the active portfolio account by account, evaluating each borrower against the current credit policy's top-up and cross-sell eligibility rules: EMI track record (number of on-time payments, bounce frequency, DPD history), outstanding principal relative to original sanction, current LTV against security where applicable, bureau score trajectory since origination, and any active EWS flags from the Early Warning Agent AI. Borrowers with active EWS flags, restructured accounts, or overdue EMIs are excluded from the eligible pool before offer generation begins — the agent does not offer credit to accounts showing stress signals.
• –Scores each eligible borrower on top-up readiness — a composite of repayment consistency, income stability indicators from the original credit file, and time-since-origination — and ranks the portfolio by conversion likelihood. Accounts in the top quartile of readiness receive immediate offer generation; accounts in the second quartile are queued for the next cycle; accounts below the threshold are flagged for monitoring without offer generation, with the expectation that continued good repayment will move them into the eligible pool over time.
• –Calculates the offer parameters for each eligible borrower: maximum top-up amount within policy limits (typically a percentage of the original sanction or a multiple of monthly income, whichever is lower per current policy), the rate band applicable to the borrower's current risk profile via the Risk-Based Pricing Agent AI, and the eligible product variants — top-up on the existing loan, a new product line where the borrower qualifies, or a tenure extension offer where that is the more appropriate intervention.

Invoked when: eligibility scoring produces a ranked eligible pool — offer generation runs immediately for the top-quartile accounts

• –Generates a personalised offer for each eligible borrower: the offer amount stated precisely (not a range), the applicable interest rate, the indicative EMI at the standard tenure for that product, the validity window (typically 30 days), and the acceptance path — the specific action the borrower takes to accept, whether that is a link to the self-serve application, a call to the relationship manager, or a WhatsApp reply. Personalised offers with specific amounts convert at substantially higher rates than generic campaign messaging; the agent generates a distinct offer for each borrower rather than a single campaign offer applied across a segment.
• –Matches the offer framing to the borrower's current situation: a borrower six EMIs into a 24-month personal loan receives a top-up framing relevant to that stage ("you have established an excellent repayment record — here is an offer based on it"); a borrower approaching loan closure receives a renewal or new product framing; a borrower whose income appears to have grown since origination (based on updated bureau data or transaction signals) receives a limit enhancement framing. The offer rationale is specific to the borrower's situation, not a generic marketing template.
• –Records the offer in the CRM with the full offer parameters, generation timestamp, eligibility basis, and expiry date — creating the audit trail for the offer from generation through acceptance or expiry. Declined or expired offers are logged with the reason (where known from borrower feedback) to feed back into conversion rate analysis and offer design improvement.

Invoked when: offer batch is generated and ready for outreach — timing model determines the dispatch schedule per borrower

• –Determines the optimal outreach moment for each borrower based on their historical engagement pattern: the day of week and time of day at which they have previously responded to WhatsApp messages, the channel through which they last made a payment (indicating active channel usage), and any upcoming repayment dates where a relevant financial touchpoint is already expected. An offer delivered at a moment when the borrower is already thinking about their loan — immediately after a smooth EMI payment, for example — converts at meaningfully higher rates than an offer sent at a random point in the repayment cycle.
• –Selects the outreach channel per borrower — WhatsApp where the borrower has high message response rates, SMS as the fallback for low-WhatsApp-engagement borrowers, email for borrowers with email-first digital behaviour, and app push notification for borrowers who are active on the lending app. Does not dispatch the same offer across all channels simultaneously; that reads as spam and damages brand trust. One channel, one offer, one clear acceptance path.
• –Monitors offer performance in real time: open rates, link click rates, and acceptance submissions per offer batch. Where a batch is performing below the historical acceptance rate baseline, flags the campaign for human review — the offer parameters, framing, or timing assumptions may need adjustment. Acceptance rates are the primary performance signal; raw dispatch volume is not a metric that the agent optimises toward.