At Global Fintech Fest 2026 in Mumbai, announcements on September 9–10 signalled a turn. India’s biggest fintechs moved from AI pilots to production. Live automation now powers customer support, fraud investigations, and in‑app search. For students and job seekers, the signal is clear. AI‑ops, risk analytics, and GenAI engineering roles are opening fast.
PhonePe disclosed new automation milestones. About 94 percent of customer support now runs on AI. Around 80 percent of fraud investigations are automated. “Not everything that is an experience today gets genuinely better using AI,” cofounder Rahul Chari said. He shared these details on September 9 during a GFF tech talk in Mumbai.

| Metric | Automation share at PhonePe |
|---|---|
| Customer support resolution through automation | ~94% |
| Fraud investigations automated | ~80% |
Razorpay showcased RAY, an AI account manager on WhatsApp, and Agent Studio to automate post‑payment workflows. Its Vulcan model targets fraud and checkout optimization across signals. Roundtables focused on moving voice AI from demo to production across functions. Topics included sales, servicing, collections, and support, signaling readiness beyond proofs of concept.
Hiring signals matched the demos. Citi’s Mumbai KYC operations opening seeks engineers to design RAG‑based workflows, agents, and human‑in‑the‑loop controls. Wells Fargo advertised a Lead Risk Analytics Consultant for GenAI. HSBC posted a business risk data scientist role with fraud analytics remit in merchant acquiring.
Priorities cluster around Python, production LLM operations, evaluation, and vector databases. Expect practical exposure to retrieval‑augmented generation, LangChain or LangGraph, tool calling, and guardrails. Understanding governance, privacy, and audit trails helps. PhonePe described “Agent Hub” automations compressing compliance reviews, highlighting process engineering alongside models.
Roundtables stressed trust as automation scales. Sessions explored identity verification, consent capture, risk controls, and responsible governance for AI‑powered engagement. Candidates who can translate RBI‑aligned policies into technical guardrails, logging, and model monitoring will stand out. Understanding consent flows and escalation paths now sits alongside pipelines and prompts.
Build one end‑to‑end mini‑system this month. Try a fraud‑signal feature store, a RAG knowledge bot, or a voice collections agent. Use synthetic data and document measurable lift, latency, and safety checks. Track employer pages and GFF exhibitor walk‑ins this week. Emphasize documentation, red‑team tests, and consent‑aware data handling.