SelfPay
Collections recovery, orchestrated before the field arrives.
SelfPay is IDAI's platform direction for retail lenders: a governed AI recovery agent for each borrower — holding continuous context that no human desk can scale — to drive self-pay first, then escalate intensity and hand residual risk to the correct next-action team: field, asset repossession, legal, or written-off recovery.
Recovery quality. Cost-to-collect. Conduct control.
GenAI operating model
One governed recovery context per borrower.
Human agents cannot hold continuous, complete context across application data, payment history, channel responses, BBPS BillFetch signals, preferred payment modes, and prior offers. SelfPay is designed so each borrower is represented by a governed AI recovery agent that can — within policy — choose timing, channel, message, and when to escalate to a human.
Context the agent holds
- Application and product profile
- Past payment and bounce pattern
- Communication response history
- BBPS BillFetch and payment-rail signals
- Preferred modes (UPI, mandate, branch, cash path)
- Prior offers, PTP outcomes, and suppress windows
Right time
Engage when propensity and contactability align — not on a fixed calendar that burns goodwill and cost.
Right channel
Voice, SMS, WhatsApp, IVR, or silent payment path based on evidence of what works for that borrower.
Right message and offer
Personalized, policy-safe content; optional auto-engagement offers only when risk and conduct rules allow.
Right human handoff
When digital exhaustion is real, assign the right-fit field or specialist agent with full context for doorstep collection, asset action, or write-off recovery.
Unlike unmanaged chatbot experiments, SelfPay agents operate inside treatment policy, audit trails, and recovery-conduct constraints. Personalization without governance is a regulatory liability.
SelfPay operating map
One governed borrower context. Multiple controlled recovery paths.
The agent is not the product story by itself. The operating system is context, policy, treatment selection, channel orchestration, and clean next-action handoff.
Borrower context
Payment behavior + response history + policy
Channels
VoiceSMSWhatsAppIVRUPI / BBPSNext action
Self-cureHuman voiceFieldAsset / legalWrite-off recoveryThe structural problem
Customer books compound. Field teams compound with them. Margins do not.
Most collections engines still lean on static rules, calendar-based campaigns, and late escalation to expensive human capacity. Early-bucket customers who would have cured on a clean digital path are treated like hard delinquents — while true risk waits in the same queue. The result is rising OpEx, uneven recovery, and mounting conduct risk under tighter recovery standards.
Typical field visit cost
~INR 200
Per successful or attempted field contact economics vary by city and agency model; order of magnitude is the point.
Digital self-pay path
~INR 20 class
Online payment rail / gateway economics plus platform fee — an order-of-magnitude structural shift when propensity is correctly selected.
What compounds
Customers × field force
As books grow, collections teams grow linearly. SelfPay attacks the part of the book that should never have required field in the first place.
The Cure Ladder
Intensity only when evidence demands it.
SelfPay uses a governed ladder from self-cure to specialist action — designed for India's UPI and BBPS-class rails, and portable to non-bank lenders wherever digital payment friction and field cost pressure coexist.
- 01
Self-cure
Lowest
Identify high-propensity bounce and early-delinquency customers. Suppress conflicting field and aggressive human contact for a defined window. Deliver simple payment paths (UPI / BBPS-class rails where available) with precise, respectful messaging.
- 02
Assisted digital
Rising
Orchestrate personalized voice, SMS, WhatsApp, and IVR with increasing intensity only when response and risk justify it. Sequence is governed to prevent conflicting or excessive contact.
- 03
Human voice
Elevated
Route remaining cases to skilled tele-collections with context, promise-to-pay discipline, and full auditability. Agents receive prioritized signals and treatment history.
- 04
Field collection
High cost
Deploy field only when digital and voice economics fail the case. Target: convert a meaningful share of visits that once cost ~INR 200 into digital recoveries near payment-rail cost (~INR 20 class) plus platform fee — a structural bargain for the lender.
- 05
Specialist next action
Specialized
Hand over cleanly to the right team: field intensification, two-wheeler or asset repossession, legal, or written-off recovery partners — with full treatment history and reason codes.
Capability spine
Good-bounce selection
Separate customers who will self-pay under the right conditions from those who need human or field intervention — using behavioral, payment-rail, and portfolio signals rather than static rules alone.
Intensity orchestration
Voice, SMS, WhatsApp, and IVR campaigns with governed escalation. Intensity rises with evidence, not with calendar habit.
Treatment governance
Suppress conflicting actions, respect conduct windows, keep audit trails, and protect borrower treatment quality under rising regulatory scrutiny.
Clean handoff
When digital exhaustion is real, route to field, asset, legal, or write-off teams with treatment history, reason codes, and prioritized context.
How institutions engage now
A product direction that can be tested before a platform-scale commitment.
CXOs are right to be wary of pure advice without operating proof. SelfPay is therefore introduced through selected design-partner diagnostics and governed pilots: a demoable operator console for executive walkthroughs, institution-specific data mapping after PII controls, and AI-assisted operating procedures under expert control until modules earn the right to harden into product.
- Confidential data preparation: PII stripped or tokenised before any analysis environment outside the institution’s control.
- Institution-specific schema mapping: every lender’s files differ; the spine is the Cure Ladder and economics, not a rigid import template.
- Demoable operator console for executive walkthroughs — strategy, segment selection, treatment preview, and outcome readout.
- Backend may begin as AI-assisted operating procedures and scripts under expert control, then harden into product modules as pilots prove value.
This is how serious recovery programs were already run in pre-genAI environments with lightweight, high-ownership operating tools — now accelerated with modern AI for analytics, segmentation, and orchestration design while accommodating each institution's data reality.
Early access
Register interest for SelfPay early access and pilot conversations.
Priority is given to regulated lenders and non-bank portfolios with measurable early-bucket volume and a named collections sponsor. Work email required.