Capri Global builds AI-led underwriting and collection systems
Key highlights
- Capri says more than 150 on-roll technology employees develop its platform in-house.
- Capri reported gross NPA of 1.2% and net NPA of 0.8% at Q2FY27.
- Capri's PCR declined to 33.3% at Q2FY27 from 41.2% at FY26.
- Insurance premium increased 43% to Rs 143.4 crore in H1FY27.
powered by IRIS

Capri Global says its 150-plus on-roll technology team is building an in-house digital lending platform spanning underwriting and collections, with AI-powered call-centre analytics and machine-learning early-warning models in its Q2FY27 presentation. The named tools include Orion LOS for approvals, Sentinel 3.0 for scoring, and Pegasus for real-time collections monitoring.
Know Capri Global Capital inside out
Results, valuation, peers and who's buying, in one view.
How large is Capri's in-house technology operation?
Capri says its dedicated on-roll technology team has more than 150 members and is committed to developing advanced solutions. The platform is developed end to end in-house, which Capri says supports low maintenance costs, rather than relying on an externally supplied lending-system workflow.
Capri says its integrated workflow and digital loan journey are designed for faster turnaround time, or TAT, while dashboards monitor performance, productivity and cost efficiency. Its system architecture is described as customizable and scalable, with the capacity to adapt to regulatory changes. The presentation also says sensitive customer data are encrypted under what Capri calls the highest information-security standards.
Capri is developing generative artificial intelligence, or generative AI, capabilities for process automation. The presentation does not provide a deployment date, user count, cost reduction, turnaround-time improvement or recovery outcome for those developing capabilities, distinguishing the disclosed plan from the already named underwriting and collection applications.
Which applications does Capri use for loan underwriting?
Capri uses its in-house Orion loan origination system, or LOS, as a single-window loan-approval workflow. The underwriting process also includes a central processing unit for document validation, a WhatsApp-integrated query module and dashboards for TAT monitoring.
Capri lists the Athena App for income and collateral assessment and geotagging, while the Daedalus App supports technical verification and collateral evaluation. Sentinel 3.0 is identified as a scorecard and business-rule engine, or BRE, and sits alongside risk-based pricing. These tools cover document review, borrower assessment, collateral checks and approval workflow rather than a single standalone AI underwriting product.
The workflow begins with four disclosed lead-generation channels. Spark Mobility App is for the direct-sales team, Leado Portal supports the tele-calling unit, and Capri Business Partner App serves direct selling agents and connectors. The Lead Management Portal aggregates multiple channels before an application reaches Orion LOS.
How does Capri use AI in collections?
Capri uses the Pegasus Collection Dashboard for real-time monitoring by collection channel, geography and individual. Its Pegasus App is designed to optimize routes, schedule field visits, track field agents live and record collection status, while incentive gamification uses a performance-driven leaderboard.
Capri also lists AI-powered call-centre analytics in its collections workflow. Its collections presentation specifies generative AI-powered speech-to-text conversion for call-centre analysis and identifies higher collection efficiency, improved resolution rates, higher non-performing asset recovery and reduced collection cost as intended outcomes. It does not quantify the effect of the call analytics on any of those measures.
Machine-learning, or ML, tools are applied before and after payment bounces. Capri's early-warning system, or EWS, profiles customer risk and predicts bounce rates and non-performing assets, or NPAs; the pre-bounce process also includes AI-based automated bot calling and pre-equated monthly instalment reminders through WhatsApp and short message service.
After a bounce, Capri says its sloppy-payer model predicts roll-back and roll-forward behaviour, and an analytics engine estimates whether a borrower will honour a promise to pay. The process uses days-past-due, or DPD, bucket-wise collection strategies. For post-delinquency accounts, Capri lists an NPA recovery model, automated legal escalation with TAT tracking, and tele-calling for high-risk pre-delinquent borrowers.
Got a follow-up? Ask IRIS
What do Capri's asset-quality figures show alongside these tools?
Capri reported gross NPA of 1.2% and net NPA of 0.8% at Q2FY27, compared with 0.9% and 0.5%, respectively, at FY26. Its provision coverage ratio, or PCR, declined to 33.3% at Q2FY27 from 41.2% at FY26. Capri also reported collection efficiency above 98.0% and said collections were managed fully in-house.
The movement from FY26 to Q2FY27 was therefore an increase of 0.3 percentage points in both gross and net NPA, while PCR fell by 7.9 percentage points. The disclosed figures provide the operating setting for Capri's EWS, sloppy-payer and recovery models, but the presentation does not attribute the change in asset quality or collection efficiency to any named technology tool.
Asset quality differed across lending segments at Q2FY27. Micro, small and medium enterprise, or MSME, loans had gross NPA of 3.1%, net NPA of 1.7% and PCR of 45.2%, compared with 1.2%, 0.8% and 37.9% for housing loans. Gold loans recorded gross NPA of 1.2%, net NPA of 1.0% and PCR of 19.1%, while construction finance reported 0.0% gross and net NPA with PCR of 50.0%.
How does Capri's digital journey continue after approval?
Capri extends its digital workflow beyond underwriting through electronic signing and electronic stamping, biometric and face authentication, and electronic National Automated Clearing House, or e-NACH, registration. These digital-disbursement mechanisms follow the Orion LOS approval workflow and precede the servicing and collections applications.
The Capri Loans App provides loan documents, service requests, payments and loan top-ups after disbursement. Capri AI-Dost is listed as an AI chatbot for customer engagement. Together with Pegasus and the collection models, the product suite covers customer contact from lead generation through servicing, repayment monitoring and delinquency handling.
Capri's insurance-distribution page shows another digital customer journey. Total insurance premium increased 43% to Rs 143.4 crore in H1FY27 from Rs 100.4 crore in H1FY26. The presentation lists real-time policy issuance through digital consent, a buy-and-compare portal, digital customer onboarding and WhatsApp links supporting a two-click buying journey.
The insurance data also show that direct distribution represented 65% of premium and cross-sell represented 35%, based on Q2FY27 premium mix. Health and wellness accounted for 56% of premium, compared with 24% for life insurance and 20% for general insurance. These insurance tools are separate from the disclosed mortgage-loan application suite, but both use digital onboarding and customer-service mechanisms.
Conclusion
Capri has disclosed an in-house technology model built around more than 150 on-roll employees and applications for sourcing, loan approval, collateral assessment, disbursement, servicing and collections. Its approach combines rule-based systems such as Orion LOS and Sentinel 3.0 with AI and ML tools for speech analysis, automated customer contact, early warnings and delinquency prediction. The Q2FY27 asset-quality data provide context for these systems but do not establish a causal result from them.
The next disclosed development to watch is Capri's plan to build generative-AI capabilities for process automation. Future presentations may show whether the company provides adoption, turnaround-time, cost, recovery or collection-efficiency measures for that plan, and whether gross NPA, net NPA and PCR change from the Q2FY27 levels of 1.2%, 0.8% and 33.3%, respectively.


