From Face Match to MLOps: How Airtel Payments Bank Is rewiring digital trust with AI

As digital transactions surge across India’s fintech ecosystem, Airtel Payments Bank, financial services arm of the telecom company Bharti Airtel, is sharpening its AI strategy with a firm focus on explainability, ethics and regulatory readiness. In an interview with TechCircle, Chief Information Officer (CIO) Pinak Chakraborty said that the bank is developing “human-centred, responsible AI systems” that accelerate innovation without compromising trust, compliance, or customer safety.

Chakraborty said that every AI-driven decision at the bank undergoes rigorous governance checks — from model version control and continuous testing to human oversight at critical stages. “Innovation and integrity must go hand in hand,” he noted, adding that transparency and explainability are mandatory before an AI model goes live.

At the core of this approach is Airtel Payments Bank’s AI governance framework, built on the pillars of fairness, accountability, safety and trust. Cross-functional roles across data science, risk, governance and compliance ensure clear ownership as models move from design to deployment. Early-stage risk assessments, continuous monitoring and periodic bias reviews help the bank maintain speed while aligning with evolving regulatory expectations.

A continuous MLOps framework supports this at scale. Today, the bank uses a combination of automation and manual checks to reduce bias, secure models, and ensure audit-readiness. As the bank expands its digital footprint, Chakraborty said automation within MLOps will deepen, but “rigorous standards for compliance and fairness will remain non-negotiable.”

One of the bank’s most impactful use cases is Face Match, an in-house AI security feature that evaluates a threat score based on user behaviour, transaction patterns, device intelligence and historical data. If the risk threshold is breached, the system triggers Face Match authentication. Users are prompted through the Airtel Thanks app to verify their identity via selfie; the image is checked against onboarding records using facial recognition and liveness detection. A successful match restores normal banking access.

This sits alongside a broader anti-fraud shield that analyses hundreds of risk signals in milliseconds and scans billions of URLs daily to block phishing links across SMS, WhatsApp, email and browsers. Together, these systems offer over 99% accurate anomaly detection, while in-app alerts and education tools give customers instant control.

Chakraborty emphasised that while AI underpins much of this, the bank’s operating model remains “deeply human-centric.” AI powers personalised recommendations, customer support, and proactive risk interventions, but is designed to reduce friction — not add complexity — for India’s diverse digital population. Features like Safe Banking alerts, Fraud Alarm and Face Authentication with liveness detection aim to make digital banking intuitive even for first-time users.

“Trust is another priority,” said Chakraborty. The bank builds explainability into every high-impact decision — whether a fraud flag or an eligibility score. Human oversight is retained for all critical cases, supported by strong data governance, continuous bias checks and customer-facing education. “Even customers new to digital banking must feel that AI is a reliable, accountable partner,” he said.

On the financial inclusion front, Chakraborty said AI is helping the bank deepen its reach across rural and semi-urban India. Predictive analytics recommend relevant products, while geo-analytics and behavioural models help optimise its extensive agent network. Alternative-data insights — from transaction histories to mobile usage patterns — support responsible credit and service decisions for customers lacking formal financial records.

With India’s DPDP Act and RBI’s FREE-AI guidelines taking shape, the bank has adopted a compliance-by-design approach, embedding consent management, data minimisation and explainability from the earliest stages of model development. Regular audits and a Model Risk Management Committee ensure alignment with evolving regulatory frameworks.

Looking ahead, Chakraborty said the next phase of AI-led banking will be hyper-local and voice-driven. Vernacular conversational AI interfaces will help ease onboarding for new users, while predictive models will anticipate the needs of small businesses, farmers and households in underserved regions. “AI will not only automate banking — it will make it more personalised, accessible and empathetic,” he said.