With the rise of GenAI and Agentic AI, where do you see the most significant opportunities for enterprises, and what risks should leaders be prepared for?
The biggest opportunity is moving AI from answering questions to getting work done. GenAI made knowledge accessible. Agentic AI lets systems plan, act across applications and close the loop on a process. The early value is in high-volume, cross-functional work such as order-to-cash, claims, supply chain exceptions and analytics, where agents can pull data, run the analysis and draft a recommendation in minutes.
Muthu Govindarajan, Partner – Data Engineering, Tiger Analytics
The risks shift accordingly. When AI only drafted content, a wrong answer was an inconvenience. When it takes actions, a wrong decision has consequences. Leaders should prepare for autonomy without accountability, agent sprawl across teams with no shared controls, and rising cost and security exposure as agents touch more systems and data.
The answer is not to slow down. It is to design for bounded autonomy: clear scopes, human checkpoints for high-impact decisions, and full traceability of what an agent did and why.
What are the biggest challenges enterprises face in moving from AI experimentation to scaled, enterprise-wide adoption?
The technology is rarely the bottleneck. Most enterprises can build a convincing pilot in weeks. What stalls them is everything around it.
First, pilots are often built in isolation, without a clear path to production; integration with core systems, security reviews and support models are left until after the pilot has succeeded. Second, a pilot runs on curated data, while production runs on the real thing: fragmented, inconsistent and governed differently in every system. Third, the operating model does not change. A model that recommends a better price creates no value if the process and roles around it stay the same. Finally, inference, monitoring and maintenance costs are underestimated, so what looked cheap as a pilot looks expensive at scale.
Scaling requires treating AI as a product, not a project: a clear owner, a defined KPI, a platform that makes the tenth use case cheaper than the first, and change management budgeted from day one.
How can organizations build the right data foundation and governance framework to become truly AI-ready?
AI-readiness starts with the decisions you want to improve, not with the data you happen to have. Working backward from priority use cases tells you which data domains matter, what quality they need and how fresh they must be.
Three foundations matter most. The first is well-defined, owned data products that are trusted and reusable. The second is a semantic layer of shared business definitions. GenAI and agents are only as reliable as the context they are given; if “revenue” means three different things, so will the AI’s answers. The third is unstructured data, which is where much of GenAI’s value sits and which most governance programs have historically ignored.
At Tiger Analytics, we apply this thinking to data engineering itself. Our own agentic AI tools now automate complex, time-consuming work such as data modeling and code generation, so engineers spend less time on repetitive build work and more on getting definitions, quality and governance right.
Governance should be embedded, not bolted on. That means lineage, access controls and quality checks built into pipelines, and an AI governance layer covering model risk, evaluation and monitoring. Done well, governance becomes an accelerator, because teams move faster when they know what is safe to use.
What will differentiate organizations that successfully become AI-driven enterprises from those that struggle to realize value from their AI investments?
The winners will not be those with the most pilots, but those most disciplined about value.
Three traits stand out. They prioritize ruthlessly: a handful of use cases tied to P&L outcomes, with leadership sponsorship and measured results, rather than a hundred scattered experiments. They build reusable foundations, including data products, platforms, and evaluation and guardrail frameworks, so each new use case is faster and cheaper than the last. And they invest in people as much as technology, redesigning roles and workflows, building AI fluency among business users, and earning the trust that makes employees actually use what is built.
This is why AI fluency is a priority for us at Tiger Analytics. We run structured training so that everyone, not just specialists, can use AI tools well and get the best results from them. Democratizing AI this way is what turns tools into outcomes.
There is also a mindset difference. Struggling organizations ask, “Where can we use AI?” Successful ones ask, “Which decisions matter most to our business, and how can AI make them better, faster or cheaper?”
How do you see AI reshaping enterprise decision-making and business models over the next 2–3 years?
I expect decision-making to move from periodic to continuous. Today, decisions on pricing, inventory and credit are often made on weekly or monthly cycles, based on reports someone has to interpret. Increasingly, AI systems will monitor signals in real time, simulate options, and recommend or execute actions within agreed limits. Humans will focus on exceptions, trade-offs and strategy.
That changes the shape of the organization. Analysts will spend less time producing dashboards and more time framing questions and challenging recommendations. Managers will oversee teams of people and agents. Decision rights will need to be defined explicitly: what an agent can decide alone, and what needs approval.
We are already seeing this in our own work. At Tiger Analytics, delivery is becoming AI-native: AI tools and reusable skill files are built into most of the tasks our teams perform today, across the full software development lifecycle, from requirements and design to development, testing and deployment.
Business models will shift too. Services that were labor-intensive can be delivered with software-like economics, and companies with proprietary data and deep domain expertise will be able to package that intelligence as new offerings. The organizations that benefit most will treat AI not as a cost-saving tool, but as a new way of running the business.
Enjoyed this interview? Now imagine yours. Write to:
jeevika@thefoundermedia.in
About The Author
