The Story
How India's Biggest Bank Deployed AI Agents at Enterprise Scale
Kedar Ravangave, Head of Marketing at Kotak Mahindra Bank and ex-Amazon leader, is one of the few executives in India actively deploying AI agents across awareness, performance, and 1-on-1 personalised marketing — at enterprise scale. His interview with Think School revealed the critical distinction between automation and AI agents, and why banks are at the forefront of the AI marketing revolution.
~80%
Cost Saving on Entry-Level Tasks
Minutes
Creative + Campaign Output Time
1-on-1
Personalisation at Scale
Multi-Agent
System Architecture
The Critical Distinctions
Concept 1
Automation vs. AI Agents
Automation follows fixed rules (“if X, do Y”). AI agents make decisions, adapt to context, and handle exceptions — like a junior employee who can think, not just execute.
Concept 2
How an AI Agent Actually Works
Agents have a goal, tools (APIs, databases, browsers), memory, and reasoning. They break the goal into subtasks, execute, evaluate, and loop — autonomously completing complex workflows.
Concept 3
Multi-Agent Systems
Complex business problems require multiple specialised agents working together — one for creative, one for audience targeting, one for compliance, one for reporting. Each handles its domain.
Concept 4
The Jobs Question
AI replaces entry-level pattern-matching tasks. It elevates people with deep subject matter expertise who can direct, audit, and improve AI outputs. Expertise becomes a multiplier, not a disadvantage.
Key Lessons Learnt
01
Automation Is Not the End Game — Agency Is
Most companies automate workflows. The frontier is building AI agents that can autonomously achieve business goals. This shift from “do this task” to “achieve this outcome” is the leap that matters.
02
AI Doesn't Replace Marketing — It Replaces Marketers Who Don't Use AI
At Kotak, AI agents create creatives and campaign outputs in minutes. The marketer's job becomes strategy, judgment, and creative direction — not execution. Skill up or be replaced.
03
Personalisation at Scale Is Now a Baseline Expectation
Banks once sent the same message to millions. AI agents now enable 1-on-1 customised messages across millions of customers simultaneously. Brands not doing this will lose to those who are.
04
Build for Humans to Supervise, Not Replace
The best multi-agent systems are designed so humans focus on the 10% of decisions that require judgment, while AI handles the 90% that are rule-based or data-driven.
How to Apply This in Your Business
Start With One Agent, One Goal
Don't try to build a multi-agent system on day one. Pick one high-volume, repetitive workflow (e.g., lead qualification, report generation) and build one agent to handle it end-to-end.
Use AI for Creative at Speed
Train your team to brief AI agents for ad creatives, social content, and email sequences. Cut production time from days to minutes. Redeploy that time to strategy and testing.
Design the Human-AI Handoff Clearly
Map every task in your workflow. Mark which steps need human judgment. Build agents to handle everything else. Review and audit weekly — agents improve when humans correct them.