Startups

Meet Karta: The AI Agent Built to Replace Complex Customer Support

• ByStartupStory     |    June 5, 2025

In a fast-evolving digital world where businesses grapple with scaling customer support without compromising quality, two IIT Bombay and IIM Ahmedabad alumni—Vijay and Abhishek—found themselves uniquely positioned to tackle this growing challenge. Their paths crossed once again not in lecture halls, but at the intersection of AI, product, and customer experience.

Vijay had been leading the Customer Experience function at Razorpay, one of India’s leading fintech companies, while Abhishek was deep into building AI systems at Walmart, focusing on applied intelligence. Their experiences were complementary: Vijay saw firsthand the cracks in existing support models in high-growth companies, and Abhishek was already deploying AI agents in the legal sector. Their collective insight shaped a bold vision—to reinvent customer support.

As Vijay recalls from his time at Razorpay, “No one had a good support experience ever. Traditional chatbots failed in complexity, tone, and trust and workflows that break everyday. That’s when we realized the opportunity: to build a deeply integrated, reasoning-first AI agent that could fully resolve user queries, not just deflect tickets.”

Thus, Karta (worker) was born— an AI employee for customer support that is as empathetic and reliable as a Zappos-style human, that aims to delight users during every interaction.

With GenAI, we finally have the technology to truly solve support.

“But great support needs more than answers—it needs a unified context. Traditional SaaS scattered enterprise data further across tools and teams, making it even harder for tech to help. GenAI changes that. For the first time, we can gather, understand, and reason over messy, unstructured data across channels and departments—giving AI the context and tools it needs to resolve issues accurately and delightfully.”

They began prototyping ideas, drawing on their overlapping experiences. Abhishek was pulled into Razorpay to help design agentic support models. That collaboration laid the groundwork for what became Yara, Karta’s flagship AI support agent, which was launched to the market in March 2025. 

Currently we are live with one of our customers, SmartStaff, and have successfully deflected 70% tickets while maintaining human CSAT.

Rohit Jain, Co–Founder Smartstaff, Ex ZoomCar, Ex OLA “We started working with Karta roughly 3 months back on this project. Loved their approach of not just plugging in the solution but actually understanding the problem statement, do iterative product enhancements and get us to a humongous 70% auto-resolution rate (without a dent on CSAT!) in such a short period of time!”

Karta isn’t just another AI agent—it’s built with proprietary memory graph and reasoning tech that goes beyond retrieval. It understands context, learns from past conversations and feedback, and handles complex, multi-step queries with precision. Karta works seamlessly across chat, voice, WhatsApp, and regional languages, integrating effortlessly with your CRMs, tools, and even spreadsheets. It’s a self-serve, ops-friendly platform that keeps getting smarter—no engineers needed.

One of the things that truly sets Karta apart is its ability to handle complex support flows. Our agents don’t just deflect basic FAQs—they resolve L2 tickets effortlessly and even tackle parts of L3 queries. As our founder puts it, “Karta is built to follow complex SOPs with precision, giving enterprises the reliability they’ve always wanted from AI.”

One of the things that stands out about our agents is the level of complexity it can handle. We can easily deflect L2 tickets and even the a portion of the complex L3 tickets. As the founder says, “Our AI Agents are optimized to follow complex SOPs, giving maximum reliability to the enterprise”. 

The journey wouldn’t have been possible without the team that drives Karta forward. Experienced (12+ years) founding IITB and IIMA Alumni — entrepreneurs, a blend of hustlers, computer scientists, PhDs, and veterans from companies like Amazon, Walmart, and Razorpay.

 

This high-agency, tightly-knit team is structured around ownership, trust, and a learning-first culture. Everyone contributes to the product, engages with users, and ships code. The emphasis on asynchronous work and open communication ensures a healthy work environment rooted in shared purpose.

Since its launch, Karta has gained traction in sectors like fintech, automating complex tasks involving KYC, refunds, onboarding, and salary advances. Their product continues to evolve with skills like document verification, automatic knowledge generation, and fraud detection. The goal? To become the default AI employee for customer-facing teams across India and beyond.

The company has already raised an angel round from top operators and is now in talks to raise its seed round from early-stage venture capital firms. These funds will accelerate their roadmap—expanding the team, improving automation depth, and entering adjacent sectors like healthcare.

Despite their achievements, the path hasn’t been without hurdles. Data readiness was a significant challenge: “Most enterprise data is tribal, broken, and unstructured,” they admit. They overcame this by building a knowledge graph and SOP-based execution engine. Trust, especially in regulated industries, was another concern—one they addressed with financial guardrails, transparent escalation layers, and robust logging from day one.

Asked what advice they have for aspiring founders, they respond with humility and clarity: “Startups are marathons in the dark. Build with deep insight into the problem. Talk to users daily. Ship relentlessly. And remember: great teams, not great ideas, win.”

In an era where automation often sacrifices the human touch, Karta is proving that AI can be both delightful and dependable—solving complex problems with empathy, efficiency, and precision.

Follow Startup Story