Five years ago AI was a buzzword, a technology that corporations across industries were silently working on or trying to implement. Today, it has become a force multiplier that is actively transforming the world around us. From manufacturing to coding, from HR to security, AI today is being used almost everywhere to automate workflows and increase efficiency – something that wasn’t entirely possible a few years back.
Five years ago, boardrooms wanted an AI pilot and test the technology’s viability. Today, those same boardrooms are asking a more difficult question: what is the return on investment (RoI)?
At a time when money has become an important topic during conversations around AI and enterprises are increasingly relying on open source models while investing heavily on cyber-security, including in context with the threats created by AI, we sat down for a conversation with Navtez Singh Bal, who is Red Hat’s Vice President and General Manager for India and South Asia to understand what Indian enterprises are asking for now, what are the bottlenecks, and if India’s AI story can outpace the rest of the world. Here are some of the most interesting snippets from our conversation.
Data protection is now a mandatory conversation
Conversations around AI are changing across industries. It’s no longer about finding more and more use cases and finding the right models to deploying those use cases. The conversation these days has shifted to more critical questions such as where the data is stored, who can access it and how it is used to train models. While talking data storage, Bal said that India’s Digital Personal Data Protection (DPDP) Act has turned it into a compliance priority.
“I would say it’s now a mandatory conversation, because if you go back and actually see what the DPDP Act and others talk about, there is a big set of safeguarding which you have to do when it comes to your personal data… There’s an angle around reasonable security safeguards which you need to put in… There is a piece around how do you actually restrict AI to specific consented purposes. There is a piece around algorithmic accountability and auditing… And this has to do with cross-border transfer restrictions and others,” he told HT Tech.
This conversation, however, is new. “This wasn’t the case a year and a half ago. But what we are finding is, as people are now really thinking about scaling AI within the organisation, these conversations become very important,” he added.
‘Rogue AI’ is the tip of the iceberg
During our conversation we talked about the recent reports of AI systems bypassing their own internal guardrails, something that has rattled end users and enterprises alike. Bal believes the industry has seen only a fraction of the problem, because the real destination for AI is not chat but autonomous agents.
“Whatever you’ve heard till now is actually, I would say, the tip of the iceberg, because we are still in the early days of AI development… And AI was never actually designed to get people to just chat… The end goal is to actually create a bunch of these virtual agents which can actually do tasks for you autonomously. That’s the holy grail everybody’s trying to reach,” he said.
He also noted that a lot of times, AI agents accessing data that they weren’t intended to happens because humans are unable to think about the possible scope of the search required by the AI agents to complete a task.
“…one thing we’ve learned from machines is that because they are very exhaustive in their search or reasoning, there are avenues which we haven’t thought about when we give them directions. That is where you have some of these issues which actually happen. My sense is we can all get better. We will actually use machines to, by the way, figure out the guardrails as well, so we will use AI for AI. That is the direction. But yes, it’s an important question. I think it’s a question which has to be thought through by every boardroom,” Bal told HT Tech.
AI: From experiments to a big token bill
Another interesting that the Red Hat executive shared with us is regarding the adoption of AI within some of the big enterprises. He told us that tech companies such as Microsoft, IBM and Red Hat are already implementing AI use cases across finance, HR, marketing, software development and testing and that cost factor around AI deployment is increasing becoming a part of the boardroom conversations.
“Most of them have actually just launched experiments in AI all over the place… Experiments are easy to start. But sooner or later, models become more reasoning-oriented. They start taking more time to think and give you an answer, which goes into token consumption, which actually translates into a big bill. And people are now beginning to ask the question: listen, am I using the right model for the right purpose or not?” he said.
The cost gap is not small. At Red Hat, he noted, the difference between the model used for basic productivity tasks and the one used for complex, first-of-its-kind code can run from 10 to 100 times.
As a result, Indian enterprises are now putting the right governance and infrastructure around AI, getting smarter about costs and making sure there is no “shadow AI” running in the background. “They want to make sure that AI is embedded inside their applications. And hence, we see a merger of application development and AI management… the same workflow which you use for writing applications is the same one which you use for AI. And the moment you do that, you start putting some corporate guardrails, enterprise guardrails around it,” Bal added.
What Indian enterprises want now: ‘Show me the money’
When asked how the Indian market has changed for Red Hat in the past two years, Bal said some things remain steady. “Red Hat Enterprise Linux, I think, continues to remain a bedrock, and the go-to operating system for any enterprise deployment. Nothing much has changed on that one.”
What has picked up is automation. “We are now seeing good uptake on the whole automation piece, because what people are now realising is that automation actually has to go hand in hand with security and AI,” he said.
He also told us that AI remains the biggest ask, but the tone of the conversation has changed. “There was a lot of enthusiasm for AI a couple of years ago, and everybody wanted to get started. Now the conversations are turning a little bit into ROI conversations. Show me the money. Why are you spending this much? Can you actually show what the output is?”.
Rising hardware cost, something that has impacted almost every aspect of the global tech industry, is adding to that pressure. “We are now seeing this level of hardware shortage and hardware prices going up, given the semiconductor issues in the world. So that’s actually led to, I think, IT costs going up disproportionately faster than the revenue growth in companies,” he added.
Sovereignty enters the boardroom
Interestingly enough, conversations around AI now are also including a new concept called ‘sovereignty’ now. Earlier, sovereignty included just storing the data locally. Now enterprises are also talking about sovereignty with respect to ownership of data and that of infrastructure.
“Sovereignty was a conversation which did not occur a couple of years ago. But now everybody wants to talk about sovereignty, both from a data perspective, from an ownership-of-infrastructure perspective, from control, from governance, from compliance perspectives… and I’m not seeing it only in India, but also across the region,” Bal said, adding that geopolitics and technology have become deeply intertwined.
Why Red Hat would ‘bet on India any day’
Global enterprises are investing aggressively in India and there are several reasons for this trend. The first is the absence of legacy baggage. “India does not actually carry the level of technical debt which some of the other markets carry. Now, that’s great, because what it does is it actually allows them to directly get onto the latest products,” he said. Many Indian companies, he explained, are skipping the virtual-machine era altogether and starting on containers before moving straight to AI.
The second is a deep embrace of open source, with the government as its biggest champion. “India as a market has actually embraced open source quite a lot. And I would say the government has also been the biggest local champion for open source, because effectively what they don’t want is to get caught in vendor lock-in and control,” Bal added.
The third and fourth are talent and temperament. “Our developers… outside of the US, maybe the second-largest developer population in the world… There’s a lot of contribution which comes from the developer community in India into open source products. And maybe the last thing I would say is, in India, the entire mindset is around how to make something work. No entrepreneur takes no for an answer.”
“These four forces are actually working for us: lack of technical debt, support for open source, a massive developer population, and an entrepreneurial zeal and spirit. And it’s a young country. You add all that together, I would say I would bet on India any day,” he said.
An 80/20 world: Open source models and India’s AI edge
When asked whether India’s lack of technical debt and its open source focus would speed up AI adoption, Bal said, “I think they will help even more.”
He acknowledged that AI is different from pure open source in two ways that India cannot easily change. “Number one, model development is a highly resource-intensive and expensive proposition. Number two, silicon, which is the accelerators needed for development, is concentrated in the hands of very few companies and very few countries.”
But the rise of open models will split the market. “80% of the work will get done by open source or open-weight models, and the balance 20% is where you will go to the frontier labs… On silicon, I think more and more choices will come up… I think where Indians will excel is in blending one and two, and then using it to solve real-world problems. And I think we will get ROI,” Bal said.
He was confident about where that leaves the country. “At least on AI implementation and AI productivity, India would actually be far ahead. And then we will build some of our own indigenous technologies. We’ll construct our own data centres. It’ll be good enough for day-to-day stuff. And you should realise that we are still in the early days of AI.”
‘We don’t lag in technology’
When asked where India still lags the world in technology adoption, Bal admitted that he doesn’t believe that India is lagging behind in technology at all.
“I don’t think we lag behind in technology adoption at all. In fact, I think Indian customers and Indian clients… want to be frugal in their innovation, but they definitely want to innovate… India is actually one of the most open markets when it comes to ideas and innovation, and the fastest. And if something works, it just works across the board. Everybody adopts it.”
“When it comes to digital technology, I don’t think we lag anywhere… We might lag in infrastructure, and roads, and all that other good stuff, but that’s fine. But in technology, I don’t think we do,” he added.
Bottlenecks to AI adoption
That does not mean the road is clear. Bal said Indian AI has moved past the pilot stage, and production brings hard choices.
“We are no longer in the piloting phase. We are in the production phase. The moment you get into the production phase, you actually have to make real-life choices around what to do on op-ex, what to do on cap-ex, what to do from a governance perspective… Whichever way we look at the market, we will actually have to have a hybrid setup. And hence, internal buildouts of data centres are needed.”
He listed three bottlenecks Indian organisations will face:
- Infrastructure: scaling up internal data centres, which requires capital many companies do not yet have.
- User adoption: “How do you do that change management? How do you actually train people to adopt it?”
- Business models: “If you’re an Indian service provider who’s going to use AI, you will have to rethink your own business model completely from scratch.”
“So, three things: capital constraints to build infrastructure; change management issues when it comes to people adoption; and rethinking your business models and systems from scratch, which requires a serious change in the mindset of the leadership,” Bal concluded.
