Leena AI rebuilt its platform twice to reach 70% ticket deflection

From chatbot to enterprise automation platform

Leena AI has grown to more than $20 million in annual recurring revenue by solving a problem its three co-founders didn't initially understand: the massive cost of enterprise support tickets. The company tripled revenue in the past year while maintaining flat headcount, a trajectory built on two complete platform rebuilds and a willingness to follow customer demand rather than founder vision.

CEO Adit Jain and co-founders Mayank Goyal and Anand Prajapati—all IIT Delhi graduates with no corporate experience when they started—launched their first company, Chatteron, as a horizontal chatbot builder in 2018. When they examined who was actually paying them, they found 15 customers using the tool for something unexpected: internal IT and HR support chatbots. That discovery triggered their first pivot to Leena AI, focused initially on HR ticket deflection.

The second pivot came in 2024 and required tearing down the entire architecture. The original system, built on BERT and NER models, could deflect only 35-45% of tickets. Large language models made true agentic automation possible, but capturing that opportunity meant rebuilding from scratch and migrating every customer. Jain described 2024 as a "painful year" that also involved replacing roughly 60% of the team.

Why it matters

Enterprise employees generate an average of 36 support tickets per year across IT and HR alone, at $20 in direct costs per ticket and up to $70 when factoring in productivity loss. Vendor-specific automation tools from ServiceNow, Workday, and SAP have failed to solve the problem because enterprises run on multiple systems that don't interoperate. Leena's vendor-neutral architecture addresses the multi-application reality of how companies actually operate, and its 70% deflection rate demonstrates the commercial viability of agentic AI in back-office operations.

Architecture designed for enterprise complexity

Leena's rebuilt platform centers on "AI Colleagues"—purpose-built Level 3 agents that handle work individual contributors currently do in HR, IT, and finance. An orchestrator evaluates incoming tasks and routes them to the appropriate agent. The system switches between frontier models like Claude and GPT depending on which performs best for a given task.

Four layers make the AI Colleagues functional in enterprise environments. The Agent Operating Protocol grounds each agent in company-specific business processes. A skills library provides more than 200 pre-built integrations with enterprise applications that customers can activate immediately. A context graph improves with each use and suggests refinements to operating procedures. For legacy systems without APIs, Leena's operating browser allows AI Colleagues to access information the way humans would.

Hallucination risk is managed by running two LLMs in parallel at every step—one makes the plan, the other checks it—reducing the error rate from roughly 2.5-3% to approximately 0.09%. A third model flags anomalies after execution and surfaces them to the customer's own process expert, not a Leena employee.

Pricing follows outcomes, not seats

Leena moved from per-employee-per-year pricing to a platform fee plus consumption model. The platform fee covers access and enough tokens to deploy the first one or two AI Colleagues. Beyond that, usage drives billing as customers automate more processes. The company has also run outcome-based deals where Leena pays back if ticket deflection falls below agreed thresholds and earns more if customers exceed them. Variable revenue currently represents less than 10% of overall ARR.

Customers are now using the platform for use cases Leena wasn't originally designed for, including end-to-end recruitment automation and accounts payable processing. One customer, Coca-Cola, reduced turnaround time across HR, IT, and finance from two days to six hours.

These details were first reported by Bessemer Venture Partners in a case study on Leena AI's evolution from chatbot to enterprise automation platform.