AI Testing Boosts Trust in BFSI, Madhav Bhadra, founder & CEO, AQM Technologies

By Puja Sharma

AI-powered testing is helping BFSI firms strengthen trust, improve compliance, and accelerate secure digital innovation. IBSi spoke with Madhav Bhadra, founder & CEO, AQM Technologies.

Given BFSI’s reliance on trust and compliance, how will AI-powered testing frameworks redefine customer confidence and regulatory adherence?

The confidence and regulatory adherence need to be viewed from three angles under AI environment. One, the LLM models and their reliability in terms of accuracy and drift. Two, the agentic platforms and frameworks that are deployed for the execution of specific tasks. Three, governance and guardrails are deployed for the regulated deployment of AI in an organisation. The contextual data availability and structuring are assumed here, since that is a foundational requirement for reliable AI-driven activities.

The AI-powered testing frameworks, such as TestBlend.AI, can help assess and ensure that AI deployment in an organisation has been done keeping all three angles in consideration.

With rising cyber threats, how should QA and cybersecurity testing converge to deliver both resilience and speed in BFSI and other regulated industries?

In the evolving AI environment, the AI deployment cannot be looked at in isolation, only for productivity improvement. Besides functional accuracy, consistency and robustness, cybersecurity becomes a critical aspect while deploying AI. The functionality and security of AI are increasingly converging. The future approach needs to be more comprehensive. Using AI for securing AI is going to be the roadmap ahead in the near future. The OWASP Top 10 for LLMs serves as a foundational benchmark for safe AI implementation, alongside other evolving frameworks and standards such as the NIST AI RMF, ISO/IEC 42001, and the EU AI Act. However, these benchmarks are only guiding frameworks. The hackers are using AI to find new vulnerabilities and exploits; organisations have to be equally agile and active in proactively tackling such emerging threats.

In an era of LLM-driven testing, how critical is data sovereignty, and what models ensure enterprises retain control without sacrificing innovation?

In a still evolving regulatory landscape around the LLM environment, the data ownership is increasingly becoming a pertinent question. The data usage can often lack transparency. Besides, the BFSI market has many more compliance obligations related to data privacy. The accountability of preserving data ownership and sanctity largely rests with data owners or custodians. The Agentic platform needs to take care of preserving data sanctity and security. The architecture of the platform should be such that the data exposure is minimum, along with proper masking in order to preserve the privacy and ownership of the original set of data. Simultaneously, it is also important to design the platform structure to reduce the cost of AI usage. The more complex the structure, the higher the token usage. Hence, Agentic platform should be able to take decision about which data can be exposed in order to reduce the cost of AI and which must be preserved irrespective of the cost of AI. In summary, the deployment models are still evolving. However, a platform which vectorises the data, provides inverted referencing (such as RAG) with right amount of data masking should be able to provide control over master copy of data without sacrificing innovation.

“Agentic automation is redefining QA from execution to orchestration. The focus is shifting toward validating intelligence, not just testing software.”

How will agentic automation reshape the role of QA teams in enterprises where traditional test automation has plateaued?

Test automation has always suffered from one of the major limitations, i.e. availability of the application which needs to be automated. The cost of automating the Nth sprint for first-time testing has been a costly exercise. This has kept the automation typically atleast one sprint behind the sprint releases. So, a dependence on manual testing for Nth sprint has been heavy. Automation was largely used for regression testing of the N-1 sprint for better ROI. With the help of AI, this dynamic is changing. Tools such as TestBlend.AI enable engineers to write automation scripts from Figma or Screen designs. The automation programs can be written even before the delivery of the sprint. The automation-first approach will increasingly become the norm, reshaping team composition and dynamics.

The second most important aspect is the QA of AI agents that the organisation will deploy. Since AI agents are developed in new technology, the technology to ensure the quality of the AI agent itself cannot be the older tools. It needs new era AI-powered QA agents, ATAs (Agent Testing Agent), for ensuring the quality of AI agents deployed by an organisation. ATA ensures that any agent deployed by an organisation passes through a holistic test of functional accuracy, consistency, robustness, multi-turns, output formats, security, behaviour of the agent under stress situations, information leakage during tool calling & interactions.

How can QA leaders balance human expertise with AI-driven agentic intelligence to achieve higher accuracy, faster release cycles, and reduced costs?

A faster release cycle with accuracy is the key to reducing cost. AI-driven platforms can generate faster results but accuracy of it is dependent on available contexts (model reference, historical test data, application-specific documents, diagrams, and other contextual material). The human-in-the-loop remains necessary to eliminate any remaining inaccuracies generated by AI-driven platforms. QA teams should use their expertise to sanitise the inputs generated by agentic platforms for interacting with AI (such as test scenarios, prompts, expected results etc.) and verify the output to control the drift experienced by agents.

What strategies can organisations adopt to scale digital testing seamlessly across mobile, cloud, and legacy systems using AI-driven platforms like TestBlend.AI?

TestBlend.AI is a codeless, agentic test automation platform. Organisations’ QA role will become more of an orchestrator and validator rather than an executor. Hence, QA will need more analytical skills and deep domain knowledge to control the AI output from drifting. Every member of the QA squad can contribute to automating testing activities, since no coding knowledge is needed. Organisations can focus on increasing the QA team’s deep domain expertise, along with analytical skills to interact with AI. Test execution efforts can be delegated to AI using TestBlend.AI, a Codeless Agentic Test Automation Platform.