Enterprises must carefully weigh the costs of new AI tools against their benefits: TechM’s Kunal Purohit

As businesses ramp up their investments in Generative Artificial Intelligence (GenAI), they often face challenges such as insufficient strategy, trust deficits, privacy issues, and a lack of skilled personnel that hinder their progress. In an exclusive interaction with TechCircle, Kunal Purohit, President – Next Gen Services at IT firm Tech Mahindra, emphasised the need to align GenAI with the overall business strategy rather than treating it as an isolated technology to get a positive outcome. He also explains why enterprises must carefully weigh the costs of new AI tools against their benefits. Edited excerpts.

TechM announced its new Gen AI framework in April to drive the adoption of AI across industries. What kind of measurable outcomes did you see from the initiative?

Tech Mahindra's Generative AI framework, dubbed ‘AI Delivered Right’, is helping enterprises realise measurable benefits in four areas – productivity, transformation, innovation, and assurance. Today, companies failing to adopt GenAI with pragmatism risk falling behind, as AI's impact extends beyond mere productivity gains. The strategy allows enterprises to achieve significant productivity gains in areas such as code generation, test case automation, and enterprise search for functions like customer support. For example, a big healthcare provider used our GenAI strategy to automate 50% more test scenarios for new rollouts, achieving greater coverage with existing resources which is crucial in healthcare.

Business transformation is another significant theme where enterprises can reimagine their business model and offer new and unique ways to service customers. For instance, a furniture retailer quickly increased market share using our GenAI strategy to update its catalogue and displays to meet demand for pink furniture inspired by the Barbie movie. In terms of innovation, multi-agent frameworks are unlocking new business models and intelligent automation use cases, accentuating human effort.

The ‘assurance’ part of the strategy validates AI functionality and security. TechM VerifAI enables responsible AI deployment with ongoing bias and security monitoring. We've deployed over 200 enterprise AI agents globally, showcasing effective AI adoption and scaling. Responsible GenAI delivery improves outcomes, speed, quality and scalable intelligence.

Are enterprises moving from a digital operating model to a more cognitive operational framework?

Yes, the shift is well underway. Enterprises are transitioning from traditional (brick-and-mortar only) to digital (online and omnichannel), and now to cognitive (data and AI-driven). In the cognitive operating model, enterprises are attempting to use data and AI to drive significant business outcomes. The cognitive operational frameworks enable data transformation programs to support the enterprise's AI and agentic AI transformation. To fast-track this movement, we recently launched TechM Orion to facilitate the building, deployment, monitoring, and scaling of agentic ecosystems.

How are GenAI and agentic AI tools improving business performance?

GenAI and agentic AI tools are transforming business performance by hyper-automating complex workflows, enhancing productivity, customer satisfaction, innovation speed, and risk management. While initial productivity improvement estimates suggested a 70% increase, real-world outcomes are often lower when integrating with existing technology. Integrating GenAI into workflows has also reduced cycle times, enhanced precision, and revealed new revenue streams. 

While the benefits of AI are evident, the investment required to achieve these outcomes remains uncertain. Enterprises must carefully evaluate the additional costs of acquiring or developing new AI tools and capabilities to assess the net benefit. There is still work needed to identify specific problem areas that need addressing and to determine the most suitable AI solution—whether it's simple automation, AI assistants, or more advanced agentic AI.

When it comes to GenAI deployment, what are your focus verticals and why?

We are seeing significant traction for AI across various sectors, including manufacturing, smart cities, financial services, retail, telecommunications, and public services. In manufacturing, AI automates quality control and predictive maintenance, reducing downtimes and improving yields. The Manufacturing Xperience Centre allows clients to prototype AI-driven Industry 4.0 innovations. A home appliance manufacturer improved multilingual workforce communication with a GenAI translation-powered mobile app.

In financial services, we developed an agentic AI-powered unified system for a technology company, providing seamless access to card perks and personalised recommendations. The system also streamlines credit risk evaluation, fraud analytics, and regulatory reporting.

Public sector and smart city initiatives use AI for urban planning, traffic prediction, and citizen service chatbots. Multi-agent systems enable governments to manage service requests autonomously, enhancing responsiveness. Across sectors, the emphasis is on outcomes like compliance automation, faster turnaround times, and cost optimisation, integrating AI into decision-making, service delivery, and value creation, particularly in the Middle East and India.

In data-sensitive sectors like financial services, healthcare and government, among others, how do you ensure adequate security measures are in place?

In data-sensitive sectors, security in AI design is paramount. A layered approach encompassing data encryption, access control, role-based governance, and audit trails is employed for all AI interactions. GenAI models are deployed in tightly governed environments, often on-premises or in hybrid clouds, based on compliance needs. Agentic AI systems undergo bias mitigation and regular evaluations for explainability and traceability. ModelOps pipelines continuously monitor AI performance post-deployment, ensuring ongoing security and compliance. Benchmarking AI models for accuracy, fairness, and robustness is increasingly common before deployment, informing architectural decisions. With regulations like the EU AI Act, financial institutions are integrating AI assurance as a core requirement for scalable trust.

How are you setting up guardrails and adopting responsible AI practices?

Responsible AI is now foundational. Enterprises embed it across the AI lifecycle, from data preparation to deployment and monitoring. Guardrails include bias detection frameworks, secure data handling, model explainability, outcome tracking, and human-in-the-loop validation. Assurance is built into AI strategies, ensuring model compliance, safety, and alignment with purpose. 

Internal Responsible AI charters and ethical checklists are widely adopted. "AI usage boundaries" are implemented within employee workflows, defining AI's permissible scope. AI governance boards and ethics review panels are becoming standard, positioning proactive businesses to scale AI responsibly, build stakeholder trust, and mitigate risks as global regulations tighten. The goal is responsible AI design and delivery at scale.

What are your AI objectives for 2025-26, including recruitment and expansion plans in India?

Our AI goals for 2025-26 are centred on accessibility, quality, and responsible implementation. We've trained over 77,000 employees in AI, and our globally recognised four-tier AI proficiency index facilitates skill advancement. AI is being integrated across the employee lifecycle, from recruitment to project execution. We aim to empower everyone with AI, balancing human and AI efforts to improve business outcomes. 

As organisations evolve into Sentient Enterprises, we support them in leveraging AI to create new value avenues, ensuring solutions are enterprise-grade, secure, and responsibly deployed with adequate guardrails. Besides, we are deeply invested in AI and Quantum technology research, with our Makers Labs playing a crucial role in advancing towards concepts such as artificial general intelligence.