Promise and reality 

As a veteran of many tech trends, I asked Dutta if he has always thought that what sometimes appeared to be a mania for AI is genuine or puffed up by industry hype.  

“There’s something real about this wave that’s both real and surreal,” he says. “I wouldn’t have imagined two or three years ago [that things would grow so quickly].” 

As an example, he cites the startling progress of reasoning models – these are, essentially, large language models (LLMs) buffered by reinforcement learning, most famously in the example of OpenAI’s GPT series. Dutta says that the swift adoption of reasoning models can “jumpstart” change, but he adheres to that principle of PPT, saying that the possibilities can be squandered if the right workflows, skills and cultures are not in place. 

“The promise of the technology is there, as can be seen in agents, but you need people, process and technology,” he adds.  

That is particularly important in enterprise AI where organisations seek to meld the power of AI and ML to corporate systems. Agents working on private data necessitates a significant retraining of people. There will need to be “fine tuning” with “the human in the loop” leading. “IT workers need to become AI workers” is a favoured refrain. 

AI everywhere 

Beyond that, there will need to be a culture of mass adoption of common tools among workers of all kinds.  

“If they don’t use AI in some way, something will be missing in understanding, and you have a situation where initially there’s a lot of euphoria and then…[nothing]. For production and data science [to succeed], you have to eat your own dog food.” But by fostering adoption, “Jevons Paradox”, where efficiency in technology prefigures a spike in demand, may strike. 

Back to the PPT triangle: as an AI leader himself, where does Dutta believe that people like himself should sit in the organisational chart? 

“The promise of the technology is there, as can be seen in agents, but you need people, process and technology”

Debojyoti Duttaas, Nutanix

He says he believes in a matrix where the AI chief works closely with the CIO and CISO but also departments such as legal, product development and engineering. “You don’t want to reinvent the wheel, but this is an early stage in enterprise AI and it suits having a virtual team,” he says. “The market is moving so fast that if you’re not tech-savvy you may not see the tectonic plates shift, so the AI lead needs to be a trusted advisor to others.” 

Despite the early-stage AI diagnosis, Dutta says there have been real results internally at Nutanix, pointing to improved systems for site reliability engineers. Lessons have been learned and shared with customers: notably, private data must be clean and governance guardrails are non-negotiable. A proof of concept can be created quickly with OpenAI or Gemini, but when the application is back on-premise, the governance structure must be in place. The real challenge is not in finding the right tools, but in matching them with clean enterprise data, working hard with data science teams and evolving accuracy over time. 

Not everything is ready now, he says, noting that AI-superpowered software development still needs human oversight and a deep understanding of prompts and complex domain logic design. But agents (and, increasingly, agentic AI) are moving on and the only sensible attitude is to foster new thinking across colleges and in the current workforce. Also, storage, networking and other infrastructure should be optimised while quantum computing can be a powerful future ally. 

In the meantime, Dutta says, we can all benefit from having “digital minions” as assistants that can help with rote tasks, spot anomalies and at least parse, summarise and create the first draft of complex documents. If we combine our enthusiasm to learn with that eternal need for triangulating human know-how, digital innovation and process engineering, then the future is bright.