Governing AI at the heart of India's lending infrastructure
Pennant Technologies integrated Zytra across the pennApps Lending Factory — bringing continuous AI governance and runtime guardrails to 14M+ loans, 100M+ annual transactions, and a customer base that includes four of India's top-10 HFCs and top-10 NBFCs.
Pennant platform
governed per year
on pennApps
rate in production
About Pennant Technologies
Pennant Technologies Private Limited has been building mission-critical software for banks and financial institutions for over two decades. The company's flagship product, pennApps Lending Factory, powers the lending operations of banks and NBFCs across India and Asia — from loan origination and credit decisioning through collections, servicing, and regulatory reporting.
Four of India's top-10 housing finance companies and four of the top-10 NBFCs run on the Pennant portfolio. Pennant's founding director and CEO, Rama Krishna Raju, brings 25+ years of building world-class BFSI products — and a clear view of what regulated banking actually demands from its technology vendors.
The challenge: AI governance at the platform layer
As AI capabilities embedded deeper into the lending workflow — credit scoring models, document extraction, fraud detection, customer-service agents — Pennant's banking customers faced a governance gap that no existing product addressed well.
The challenge was structural. Model risk management frameworks, designed for statistical models, weren't built for generative AI. Runtime safety tools, designed for consumer applications, weren't built for BFSI regulatory environments. And compliance tooling, designed for documentation, wasn't built to produce the continuous evidence trails that RBI examiners increasingly expect.
- Banking customers had AI systems in production with no unified inventory or consistent risk tiering — making RBI MRM compliance a manual, error-prone exercise before every inspection.
- GenAI components in customer-facing workflows had no runtime guardrails calibrated to BFSI threat categories — the risk of financial advice violations, PII exfiltration, and regulatory breach was unquantified.
- Evidence required by RBI's 2024 AI Circular — hash-chained audit trails, model approvals, control effectiveness records — had to be assembled manually and inconsistently.
- No single product addressed all three layers: governance, runtime safety, and regulator-ready evidence — for the specific regulatory environment of Indian banking.
"Banking-grade AI governance doesn't exist off the shelf. The tools designed for enterprise governance were built for compliance documentation, not for the kind of continuous evidence trail an RBI examiner actually wants to see. We needed something built for this environment from the ground up."— Rama Krishna Raju, Founding Director & CEO, Pennant Technologies
The solution: Zytra integrated across pennApps
Pennant partnered with Zytra as its founding platform partner, integrating Zytra's AI governance stack across the pennApps Lending Factory. The integration operates at three layers.
Governance layer. Every AI system deployed within pennApps — credit models, document extraction pipelines, fraud classifiers, generative AI components — is catalogued in Zytra's AI asset registry and assigned a risk tier against RBI's 2024 AI Circular framework. Validation records, control assignments, and model approvals flow into a hash-chained evidence ledger. Banking customers can produce examination-ready evidence packs without manual assembly.
Runtime layer. Zytra's Aval guardrails operate inline on every AI inference within customer-facing workflows — classifying requests across 22 BFSI threat categories including prompt injection, financial advice violations, PII exfiltration, and regulatory breach. At 11.6ms inference latency and 0.5% false-positive rate, Aval blocks real attacks without blocking real customers.
Evidence layer. Every Aval decision and every governance workflow produces a timestamped, hash-chained record exportable as PDF, OSCAL, or JSON. When an RBI examiner asks for evidence that a specific AI system was governed continuously — not just at point-in-time validation — the record is already there.
Results
Pennant's banking customers moved from manual, point-in-time governance to continuous, automated evidence production — with examination-ready packs available on demand rather than assembled under deadline. The AI risk register that previously required weeks of coordination before an RBI inspection is now maintained automatically.
"The question banks ask before an RBI inspection isn't 'do we have governance?' — it's 'can we prove it?' Zytra changed the answer from 'we think so' to 'here is the evidence pack'."— Rama Krishna Raju, Founding Director & CEO, Pennant Technologies
What's next
Pennant and Zytra are extending the integration to cover agentic AI deployments within pennApps — including autonomous agents used in collections workflows and multi-step document processing pipelines. These deployments carry the highest risk profile under RBI's 2024 AI Circular and require governance frameworks that extend beyond traditional MRM.
The joint roadmap includes Zytra's Agent Control Room for real-time monitoring of agentic workflows, and Pramaan for re-performable audit trails of multi-step agent decisions — enabling banking customers to demonstrate that complex AI workflows produced compliant outcomes at the step level, not just the output level.
See how Zytra fits your AI stack
30 minutes. No deck. We'll map your AI risk surface and show you what governed AI looks like in your environment.