New Delhi, May 6 -- As India's digital payments ecosystem matures, digital payments company PhonePe is quietly re-architecting itself into an AI-first organisation-one that prioritises resilience, governance and long-term efficiency over speed. For Rahul Chari, its co-founder and Chief Technology Officer (CTO), the strategy is clear: don't rush AI adoption-engineer it right.

"Foundations come before scale," Chari says. "Before a single AI tool is rolled out at scale, we ensure the engineering around it-data access, authorization, auditability, and observability-is rock solid."

This philosophy-what the company calls "organisational scaffolding"-is shaping how AI is embedded across functions, from engineering to fraud detection and customer experience.

From experimentation to production AI

Launched in August 2016, the Bengaluru-based firm has grown to become one of India's most popular "super apps" for payments and financial services. At present, it is a market leader in UPI transactions, processing over 150 million daily transactions and serving over 600 million registered users.

PhonePe's AI journey has moved decisively beyond pilots. Today, it operates across three core pillars: software development, internal operations and consumer-facing products.

The most immediate gains are visible in engineering. According to Chari, AI-led interventions have driven efficiency improvements in the "high 20% range," transforming how teams approach code migration, testing and quality assurance.

"This isn't just about code assistance; it is integrated across the SDLC-from unit testing to QA lifecycle," he notes. "By automating routine tasks, we allow engineers to focus on higher-order problem solving."

Internally, the company has built an 'Agent Hub'-a self-serve marketplace of AI agents that streamline workflows and enable knowledge sharing across teams. On the consumer side, AI is increasingly visible through features like natural language-based payments, allowing users to execute complex tasks via simple voice or text commands.

AI and trust: A fintech balancing act

In a high-stakes payments environment, AI deployment is as much about trust as it is about efficiency. PhonePe is using AI to significantly rewire its fraud detection and resolution systems-without removing human oversight.

One of the most notable outcomes was that nearly 80% of fraud investigations are now completed end-to-end using AI. "Investigations into fraudulent activity are inherently complex," Chari says. "By automating the majority of cases, we've shortened resolution times and enabled our teams to focus on identifying new and emerging threat patterns."

This shift is helping the company move from reactive to proactive risk management, with AI surfacing "lead indicators" that can predict potential failures or threats before they escalate.

Human-in-the-loop, by design

Despite deep automation, PhonePe is deliberate about keeping humans in control of critical decisions. AI acts as an assistant-not a replacement.

"We maintain a strict human-in-the-loop mandate for high-stakes actions," Chari explains. "Every AI-assisted code merge requires final approval from a human engineer."

The same principle extends to security and threat modelling, where AI functions as a "challenger"-stress-testing systems and identifying vulnerabilities, while human experts make final calls.

Building the AI stack: Hybrid and in-house

At scale, infrastructure becomes the real differentiator, believes Chari. PhonePe is taking a hybrid "build-and-buy" approach to its AI stack. Core components-such as its LLM gateway and vector databases-are built in-house to tightly integrate with proprietary data systems and authorisation frameworks. At the same time, the company uses a mix of open-source and cloud-based models, allocating workloads based on sensitivity, performance and cost.

Crucially, sensitive financial data remains within its controlled environment. "Using AI the right way is better than simply being fast," Chari says. "By securing the foundation first, we achieve much higher velocity in bringing AI-powered features to users."

The road ahead: Efficiency as the new ROI

For PhonePe, AI success is currently measured through a pragmatic lens-efficiency. In a hyper-growth phase, the goal is to amplify human productivity across the organisation.

"We see AI as a tool to enhance the daily work lives of every individual," Chari says. "The real differentiator for companies will be their ability to translate efficiency gains into long-term revenue and profitability."

As enterprises move from AI experimentation to enterprise-wide deployment, PhonePe's approach offers a counterpoint to the industry's speed-first narrative: in fintech, at least, the winners may be those who build slower-but scale stronger.

Published by HT Digital Content Services with permission from TechCircle.