Testing the AI Agentic way

Conventional testing has its pros and cons. People, process, and tools can be well defined, documented, and implemented, but challenges around test data, environments, coverage, automation, and speed of testing can pop up. To address some of these challenges, automating the feasible steps using generative AI agents, is a great idea. These agents can be considered as virtual peers that can help accelerate tasks around the SDLC. They generate content for review rather than a human creating it from scratch. For example, agents can leverage user story inputs to build elaborate test cases using techniques like boundary conditions, positive/negative scenarios, equivalence partnering etc.

These AI agents are typically built with an interfacing layer that leverages techniques like Retrieval Augmented Generation (RAG) to harness the power of Large Language Models (LLMs), thus delivering useful outputs. The recent evolution of the LLMs and effective prompt engineering has further made outputs more predictable, usable, and with minimal hallucinations. Putting humans in the middle to review the steps is key to ensure better predictability and ownership.

These automation agents can easily reduce the creation time by 20 - 50% and improve coverage and quality, thus reducing prod defects and the overall cost of testing.

Paper | Presentation

Praveen Bagare

Praveen Bagare is a seasoned IT leader with over 21 years of experience in delivering innovative AI-powered solutions. His expertise encompasses Program Management, Artificial Intelligence solutions, and Quality Assurance, with a specialization in Test Data Management (TDM). Passionate about solving complex challenges, Praveen drives excellence and innovation. Currently, he leads EPAM’s North American TDM Competency Center, overseeing implementations that leverage generative AI and other TDM tools. He has architected, implemented, and managed TDM solutions for Fortune 500 companies across the United States.

Praveen is a recognized authority in the field, having authored a white paper on TDM in the Software Testing Life Cycle and another on AI and ML in TDM in PNSQC. An active senior IEEE member, TREWS fellow member, and presenter at AI Con USA 2025, he is committed to advancing industry standards.

When not leading teams or driving innovation, Praveen enjoys spending time outdoors with his family in Wisconsin.