How Kusho AI Wants AI Agents to Catch Software Bugs Before Users Do
As software becomes increasingly complex, writing code is no longer the biggest challenge. Maintaining it is.
Modern applications depend on APIs, intricate workflows, and third-party integrations that must continue functioning seamlessly every time new features are released. While artificial intelligence has made it easier for developers to generate code, engineering teams still spend a disproportionate amount of time testing releases, hunting for bugs, and ensuring that updates do not break existing systems. This maintenance burden slows innovation and increases the risk of costly outages.
Founded in 2023 by Abhishek Saikia and Sourabh Gawande, Bengaluru-based Kusho AI is building an AI-powered platform designed to automate much of that work. Its AI agents generate, run, and continuously update tests for APIs, web applications, and mobile apps. The goal is straightforward: free engineering teams from repetitive maintenance so they can focus on building new products.
From Product Management Frustration to AI Testing Platform
The idea for Kusho AI grew out of Abhishek Saikia’s experience as a product manager at Flipkart. A graduate in economics and chemical engineering from BITS Pilani, Saikia repeatedly saw how testing and quality assurance consumed valuable engineering bandwidth. He later partnered with his college batchmate Sourabh Gawande, who serves as Co-founder and CTO. Gawande previously worked as a full-stack engineer at FalconX.
Together they set out to create a system that behaves more like an AI software engineer than a traditional testing tool. Using proprietary models layered on top of foundational large language models, Kusho AI tests code both during development and after every release, catching issues before they reach production.
How Kusho AI Works in Practice
Consider a food delivery app launching a new recommendation engine. The feature must work correctly for users with five previous orders as well as those with thousands, across multiple cities and usage patterns. Manually designing test cases, writing the necessary code, running the tests, and reviewing results can take hours or even days—and still miss critical edge cases.
Kusho AI automates the entire process. The platform first builds an understanding of a company’s software environment. It then simulates how a human tester would approach the problem: identifying important scenarios, generating the required tests, executing them automatically, and flagging issues that need attention.
Saikia says the process runs “almost 100x more efficiently and at a much faster pace” than manual testing.
The platform supports teams across the full software release cycle. Before code is merged, it automatically generates and runs tests to surface risks early. At release time, it acts as an additional checkpoint, verifying whether the new code is ready for production. Saikia describes this as one of the company’s strongest advantages for enterprise customers: “Even if your developer is in a hurry and wants to quickly push to production, Kusho AI is standing there almost like a sentry.”
If a release is not ready, the system returns it with a detailed report outlining what needs to be fixed. After deployment, Kusho AI continues monitoring for anomalies and breaking changes. As the application evolves, the platform updates its own tests accordingly, creating what Saikia calls “self-healing systems that identify changes and make changes accordingly.” Every release also receives a risk score based on the underlying code changes and the tests that were executed.
Beyond individual test runs, Kusho AI positions itself as an enterprise testing control panel. It offers release gating, audit trails, domain-aware automation, and integration with existing software release workflows, giving engineering leaders visibility across the entire testing process.
Built for Environments Where Downtime Is Not an Option
Rather than relying on a single AI model, Kusho AI uses what Saikia describes as an “agentic orchestration layer.” Different testing requests are routed to specialised models depending on the nature of the application. A banking API is handled differently from an ecommerce platform, for example.
The company’s competitive advantage, according to Saikia, lies in this orchestration layer and the proprietary data used to fine-tune it. “That’s a proprietary learning that you can only have once thousands of people use your product.”
Kusho AI is deliberate about its ideal customer profile. “Our entire ICP is people who cannot afford even one minute of downtime,” Saikia says. The company focuses on sectors such as banking and ride-hailing, where even brief outages carry immediate business consequences.
The platform already serves more than 5,200 clients, including Ultrahuman, Paytm, Ola Electric, HCL, Michelin, HP, Razorpay, Roche, Schneider Electric, and Paymentus. On the free tier, it supports around 35,000 developers and QA professionals and has generated and executed nearly 10 million tests.
Product-Led Growth and Early Traction
Kusho AI has largely followed a product-led growth strategy. The platform is free to use without requiring payment details, allowing it to spread organically through developer and quality assurance communities via technical content rather than heavy paid marketing. Free users can opt in to let the system learn from their usage patterns. The company tracks which auto-generated tests users keep or discard, feeding that anonymised data back into the product to improve future recommendations.
As adoption grows within engineering teams, enterprise conversations often follow. Reported results include an Indian payments platform that reduced API testing time by 95 percent and a private bank that lowered annual QA costs by 80-90 percent. Users also highlight meaningful reductions in manual UI testing through continuous test generation.
Funding, Team, and Roadmap
Kusho AI has raised $600,000 in pre-seed funding led by Antler India, with participation from Blume Founders Fund, UpSparks Capital, and angel investors including Ultrahuman co-founders Mohit Kumar and Vatsal Singhal, and IDfy CEO Ashok Hariharan. Saikia says the company plans to raise another round later this year or early next year.
The startup is incorporated in the United States with its official headquarters planned for San Francisco. Its seven-member team currently operates from Bengaluru, with a physical San Francisco office expected in late 2026 or early 2027 as the company expands its US customer base, which already accounts for 60-70 percent of its clients.
Annual recurring revenue is in the “late six figures” in dollars and is growing between 10 and 20 percent month on month. Over the next year, Kusho AI plans to deepen its presence across banking and financial services, logistics and supply chain, and healthtech, targeting the ten largest companies in each sector.
The Long-Term Vision
For Saikia, the ambition is clear: make software maintenance so automated that engineers can focus entirely on creating new products. In a world where applications grow more interconnected and the cost of downtime continues to rise, tools that act as reliable, self-updating testing partners are becoming essential infrastructure.
Kusho AI is betting that AI agents capable of understanding complex systems, generating relevant tests, and healing themselves over time will become a standard part of modern engineering workflows. If the early results and customer roster are any indication, that bet is already resonating with teams that cannot afford to get maintenance wrong.