Agentic AI startup Karta AI has raised $3 Mn (around ₹28 Cr) in a seed funding round led by PeerCapital and Cedar Hill Capital. The round also saw participation from Wyser VC, CDM Capital and Seedstars.
The startup intends to use the fresh capital to expand its go-to-market (GTM) strategy and distribution over the next two years. A portion will also help it expand its product development, particularly capabilities around emerging AI technologies such as self learning, cofounder and CEO Vijaypal Singh told Inc42.
Founded in 2024 by former Razorpay executives Vijaypal Singh and Abhishek Kushwaha, Karta AI offers an omnichannel customer communication platform spanning voice, chat, email and WhatsApp.
The startup focuses primarily on voice-based interactions, targeting complex inbound queries that require coordination across multiple enterprise systems rather than routine frequently asked questions.
It claims to handle around 2 Mn customer conversations a month across more than 25 languages. According to Singh, it has 10-12 paying customers, including PhonePe, Mindlabs, Bajaj and Aavas.
Karta AI deploys its AI agents across customer servicing and retention workflows at financial institutions.
The startup claims its agent resolves around 70% of borrower queries end to end without human intervention, reducing the cost per query by approximately 50%. These queries include requests for loan disbursal updates, statements of account, foreclosure statements and the release of property documents.
It focuses on inbound interactions, where customers contact businesses to resolve issues such as failed payments, blocked cards, loan-related queries and insurance claims. Unlike outbound calls, which businesses typically initiate for a specific purpose, inbound conversations can involve multiple queries and require AI agents to retrieve information and execute actions across several systems.
Singh said he and Kushwaha began working on Karta AI after leaving fintech major Razorpay, where they had worked on internal AI automation. The startup initially focused on chat-based automation before expanding into voice in August 2025.
Karta AI targets level 2 (L2) and level 3 (L3) queries, which require AI agents to interpret multiple customer intents, retrieve information from different systems and execute workflows. This differs from level 1 (L1) queries, such as basic FAQs or requests that can be resolved through a single API call.
For instance, a borrower may call a lender to check loan disbursal status and request a foreclosure statement in the same conversation. The agent must identify both intents, retrieve the relevant information and initiate the required workflows.
Singh said Karta AI has achieved 70-80% end-to-end automation in its inbound proof-of-concept (PoC) deployments and has won all such PoCs so far.
The focus on complex queries reflects their disproportionate cost. According to Singh, L1 queries account for 30-40% of customer service volumes but only around 10% of costs. L2 queries contribute another 30-40% of volumes, while L3 queries make up the remaining 10-20%. Together, he estimates, L2 and L3 queries account for around 80% of customer service costs.
The platform does not automate every interaction. Queries involving fraud or other high-risk matters, along with processes lacking clearly defined standard operating procedures (SOPs), are routed to human agents. Customers can also request a human representative.
How Karta AI’s AI Tech Works
Karta AI’s proprietary context engine is designed to help its agents manage multi-intent conversations. While conventional voice AI systems typically convert speech to text, process the query through a large language model (LLM) and generate a spoken response, Karta AI adds a layer that identifies customer intent and retrieves relevant enterprise data before generating an answer.
The engine draws on SOPs, knowledge bases, previous conversations and API data. This allows an agent to handle follow-up requests within the same call without treating each query in isolation.
Karta AI integrates with enterprise systems to execute customer-facing and back-office workflows. It uses models from OpenAI and Google’s Gemini family, alongside open-source models such as Kimi and Gemma, Singh said.
Its voice platform supports around 10-12 Indian languages, while its chat offering supports more than 200 languages. The voice system is designed to accommodate Hindi-English code-switching, regional accents and dialects, though performance varies across languages.
Given its focus on financial services, the startup has built compliance measures and guardrails into the platform. Singh said agents are tested by Karta AI’s team, customers and AI-based testing systems before deployment.
Karta AI primarily serves Indian businesses in banking, financial services and insurance (BFSI), where high customer interaction volumes and complex back-office processes offer opportunities for automation. It plans to develop reusable, industry-specific capabilities before expanding into international markets.
The post Karta AI Raises $3 Mn To Automate Inbound Calls For BFSI Players appeared first on Inc42 Media.