Can AI Prevent Enterprise System Failures? A Bengaluru Startup Thinks So

When apps crash or transactions fail, it may feel like a minor inconvenience, but their macroeconomic toll accumulates over time. According to a 2026 Oxford Economics study, IT downtime costs Fortune Global 500 companies roughly 11% of their total revenue. A LexisNexis research estimated that failed payments account for 8–11% of revenue leakage across global e-commerce players.

Bengaluru-based Kusho AI, headquartered in San Francisco, is focused on addressing this problem by building what it describes as a reliability layer for modern software.

Founded in 2022 by former Flipkart engineer Abhishek Saikia and his BITS Pilani batchmate  Sourabh Gawande, the 8-member team uses AI to automate testing, observability, and system monitoring for enterprise applications. The idea emerged from the founder’s experience working on large-scale systems at Flipkart, where even minor issues could affect millions of users.

“Building things is easy. Keeping them running is the hard part,” he remarks.

Testing En Masse

Kusho AI focuses on improving how engineering teams test and monitor software systems. Traditionally, teams spend significant time manually simulating use cases, identifying potential failures, and ensuring applications function correctly across different scenarios.

The startup’s platform replaces much of this manual effort with AI-driven automation. It generates and executes large numbers of test scenarios, helping teams identify issues before they affect end users.

For example, on ride-hailing platforms, a single user action like booking a cab triggers multiple backend processes, including location detection, driver allocation, pricing, and payment validation. Testing these interactions manually can be time-intensive and may miss some edge cases.

Kusho’s system is designed to simulate these workflows at scale. According to the company, tasks that might traditionally involve testing a limited set of scenarios can be expanded to thousands or more using automated models, improving both speed and coverage.

Enterprise Use Cases

The company’s approach is particularly relevant in sectors where system reliability is critical, such as banking and financial services. Kusho says its platform is being used in enterprise environments, including netbanking systems that process high-value transactions.

While the company has not publicly disclosed its BFSI clients, it says its technology is deployed in systems where even minor failures can lead to financial or regulatory consequences.

“At 360 ONE, reliability is critical given the scale and sensitivity of the systems we run. KushoAI helped us move from fragmented testing to a more comprehensive, automated approach across APIs and workflows. What stood out was how quickly we could build meaningful test coverage without adding overhead to our engineering teams,” Santoshi Kittur, the former CTO of 360 ONE, tells AIM.

Tech Stack

Kusho builds on top of existing large language models such as GPT, Claude, and Gemini, layering proprietary fine-tuning and orchestration capabilities for enterprise reliability workflows. The company fine-tunes these models using anonymised data to improve performance in reliability-related tasks.

In addition, it uses what it describes as an agentic orchestration layer, where multiple models work together to produce more accurate outputs. This approach allows the platform to adapt to specific enterprise workflows and improve performance across different use cases.

Kusho sees its fine-tuning and orchestration capabilities as key differentiators from competitors, including Testsigma, Postman, Devzery, Roost.ai, and others.

Distribution, Funding and Growth

The startup combines a product-led approach with enterprise sales. It offers a free version of its platform to developers, allowing individual users to adopt the product within their teams.

Over time, the company hopes this bottom-up adoption can lead to enterprise-wide deployments. It says a majority of its enterprise pipeline has emerged through this user-led adoption model.

This hybrid strategy allows Kusho to reach both individual developers and enterprises. 

In 2024, Kusho AI raised a $600,000 pre-seed round led by Antler, with participation from Blume Ventures and angel investors.

The company, while not disclosing its revenue, says it has turned cash-flow positive. It is now planning a Series A raise later this year.

One of the challenges Kusho has encountered relates to how enterprises perceive AI. In its early stages, the company faced scepticism about the need for AI in reliability workflows. As AI adoption increased, concerns shifted toward its potential impact on jobs.

The company positions its platform as a tool to augment engineering teams rather than replace them. “Without AI, people were able to do X. With AI, people will be able to do 5X,” Saikia says, adding that the goal is to help engineering teams handle a larger volume of reliability tasks while focusing on more complex issues.