Exclusive: Informatica's Krish Vitaldevara on AI’s next evolution
When it comes to artificial intelligence, Informatica's Krish Vitaldevara believes the future belongs to those who get their data in order.
The Executive Vice President and Chief Product Officer at Informatica sat down with TechDay during his first Sydney visit since joining the company earlier this year, where he spoke about AI readiness, the rise of agentic AI, and how businesses can prepare for the next wave of intelligent automation.
"This is actually my first time to Sydney with Informatica," he said. "I've been here multiple times before, and it's one of my favourite places. We kicked off the Informatica World Tour Australia in Sydney recently with more than 100 attendees - it's been fantastic."
Vitaldevara, who leads Informatica's product management, customer engagement, market expansion and strategic partnerships globally, said his role is "to help the teams build what the customers are asking for".
Meeting customers in the Asia-Pacific region, he added, offers "a different perspective" from those in the US and Europe.
"The challenges are somewhat the same, but the perspectives are always different," he explained. "Each country is going through its own set of things they're optimising for."
What is the AI ROI struggle?
Vitaldevara noted that despite the global surge in enterprise AI spending, many organisations are still grappling with results.
"If you've seen the recent MIT study, more than US$40 billion has been spent on enterprise AI this year, and 95% of organisations have struggled to justify the ROI," he said. "Getting their data in order is incredibly important to achieve AI success."
He added that many businesses rushed into AI projects without first ensuring "data reliability" or "trusted compliance sources." Without this foundation, he said, "it's very hard to get the ROI."
Data hygiene and governance at the core
The CPO pointed out that industries like healthcare, life sciences and banking - where regulation enforces strong data hygiene - are leading the way. Others, he said, "have not invested enough in unifying their data silos or ensuring the data being used is of high quality."
In one example from Australia, a retail customer told him that data governance was once seen as "a tax" but is now recognised as "an impediment to extracting value."
"It turns out now there's an ROI you can attach to it," Vitaldevara said. "All of a sudden, there's this renewed yes to go fix the problems."
Building "AI-happy" data
Asked how data readiness affects AI outcomes, Vitaldevara recounted a conversation with a major pharmaceutical company that used Informatica's Intelligent Data Management Cloud (IDMC) platform.
"They told me we help them transform raw data into what they call 'AI-happy data'," he said. "AI-happy data is intelligent, contextual, governed, compliant and trusted."
He added that having a strong AI data foundation is now "a requirement for you to be successful in having any agentic enterprise - you cannot have one without the other."
From generative to agentic AI
Vitaldevara said the AI landscape has evolved rapidly from generative models to agentic systems that can reason and act autonomously.
"When generative AI came about two years ago, it was mostly about natural language interactions," he explained. "Then came copilots - AI helping with specific tasks. Now the transformation has been to more agentic systems, which is kind of the Holy Grail."
He described the shift as moving from "level one automation", which handles simple rule-based tasks, to "level two and three automation", where AI agents combine reasoning and context to make decisions.
"Agentic AI lets you use the power of models to combine different tasks and reason against them," he said. "It's not about replacing humans but adding a resource pool to the enterprise."
AI-ready data: the new foundation
Vitaldevara was clear that even the most advanced agents rely on quality inputs. A strong AI data foundation is critical for organisations' success in their AI journeys. "If your data is of poor quality, you're going to get results of poor quality," he said. "These agents need underlying data to make decisions. If your data isn't governed or is scattered across silos, you won't extract the right value."
For organisations wondering where to begin, he outlined three priorities: ensuring data quality and compliance, building semantic models that establish relationships between entities, and orchestrating multiple AI agents effectively.
"It's no longer about having a single agent," he said. "It's a collection of agents working together. Start with the outcome, then figure out the guardrails - because what good is a guardrail if you don't get to the outcome?"
Measuring AI impact
While it's still early days, Vitaldevara pointed to measurable productivity gains in engineering teams as one clear success story.
"When you look at engineering productivity, the numbers are in the 15 to 30% range," he said. "That's a huge amount of productivity lift."
He predicted that in time, enterprises will treat their custom-built AI agents as proprietary assets. "Agents are going to be your intellectual property," he said. "The real secret sauce will be how you combine purpose-built agents with your data and domain expertise."
Lessons from past tech shifts
Looking ahead, Vitaldevara expects organisations to experience a familiar cycle.
"There's going to be a proliferation - agents everywhere - and then you'll figure out which ones really translate into outcomes," he said. "Users will move from chasing value to managing cost, just like we all did with the cloud."
He also warned that governance and security would become key concerns. "Once you have all these agents, how do you make sure they're compliant and secure?" he said.
Vitaldevara noted that even Informatica went through this learning curve internally when developing its own AI engine, CLAIRE.
"We tried models everywhere," he said. "Now we're standardising so we can control them better. It's the exact same journey our customers will go through too."