“Despite all the excitement around AI, people aren’t paying enough attention to the digital foundation that is required for it to scale effectively,” said A.S. Lakshminarayanan, MD & CEO, Tata Communications at a press briefing in Mumbai.
Standing in front of a Powerpoint parade of buzzwords, he was making a rare, honest admission about the current state of the digital enterprise stack. How it’s like a Jenga tower that’s teetering on the edge of the living room table, with AI as the new clumsy kid flirting with disaster while trying to win the game.
In his opening keynote, A.S. Lakshminarayanan framed the next wave of AI adoption as a stress test for everything enterprises already duct-taped together over the last decade.
“What happened in the enterprise world before AI is that transactions scaled with humans making decisions. In the world of AI, it’s the AI that’s going to make more decisions at a faster cadence, and AI is going to help scale those decisions as well,” said A.S. Lakshminarayan or Lakshmi, as he’s fondly called.
This ultimately explains why Tata Communications is pitching infrastructure as the real AI battleground. The company’s argument is that if machine-to-machine decisioning becomes normal very soon, especially with agentic AI workflows, the east-west traffic between clouds, devices, and services becomes a multiplier problem.
In order to address this growing digital reality, Tata Communications rolled out what it’s calling an AI-ready suite. It’s essentially three platforms meant to help enterprises deploy and scale AI with what the company kept returning to as the three essentials: confidence, control, and clarity. The suite stitches together a multi-cloud network layer, an edge distribution layer, and an orchestration layer (called ThreadSpan), all built on the company’s broader “Digital Fabric” narrative which combines connectivity + cloud + edge + security as a unified architectural foundation.
But the real story wasn’t the product list. It was the context.
Lakshmi gave a crisp example from voice AI to further illustrate what’s happening. “In every voice AI interaction, the voice AI is making a number of tool calls and API calls back and forth. It’s what we call in the network world as east-west traffic of communication between machines between clouds, and that is really going to explode,” he underlined in a calm and measured tone.
Modern digital infrastructure is still “an assembly of several OEM-based technologies and islands,” according to Lakshmi, which already exhibit three key pain points: 1) Infrastructure fragmentation, 2) Costs that spike as complexity spikes, and 3) Security and governance that can’t keep up with machine speed.
“So the key thing to remember here is, AI is going to expose the vulnerabilities of islands of technologies and domains,” Lakshmi said. “Once you infuse AI and people start using AI at scale, it’s truly going to test the limits of these digital enterprises, and it can no longer manage their infrastructure as disparate islands.”
Tata Communications is essentially saying to large enterprises that before you teach machines to think faster, make sure your systems can survive that speed. And they are offering a solution to make distributed AI environments run like a system, not a collection of mismatched dashboards.
In plain English, Tata Communications’ AI suite will help enterprises connect better, run faster, while not losing control. The company formally positions the suite as three complementary offerings: IZO+ Multi Cloud Network (MCN), an Edge Distribution Platform, and ThreadSpan – all tied into the Digital Fabric approach.
Multi-cloud networking becomes complex because every cloud has different nuances. But the MCN “abstracts all the complexity, it makes multi-cloud networks work as though you are dealing with one cloud,” suggested Lakshmi. Edge matters in AI workloads because latency and security are now user experience problems. The “always-on” security angle is especially critical when AI-assisted attacks can scale too. “With AI, DDoS attacks can really suck up all your compute power and make it not work,” warned Lakshmi.
Also read: Hackers trick ChatGPT and Grok to install malware onto devices: Report
But ThreadSpan is where the penny really dropped, because it’s the piece that claims to turn the chaos of AI orchestration into something operable.
ThreadSpan is presented as the orchestration and control layer, the pièce de résistance that none of Tata Communications’ competitors has, according to their CEO. Beyond just visibility, ThreadSpan introduces agentic workflows both at the edge and across hybrid clouds, making all the pieces work together like one overarching system.
“ThreadSpan reduces the complexity, so you can not only do visibility and observability across these domains, but you can also configure and manage,” noted the CEO of Tata Communications.
That “configure and manage” bit is the real differentiator, according to Tata Communications representatives I spoke with. Plenty of enterprise tools out there will show you your mess in 4K, fewer will actually help you fix it, across vendors, without a week or more of anxiety.
Natarajan Sivasamban, EVP & Global Operations Head at Tata Communications, didn’t sugarcoat the modern enterprise stack. “If you look at any large enterprise today, it will typically be a museum of tools and platforms,” suggested Sivasamban.
“Such numbers of tools and platforms creates complexity in terms of managing… and security is the biggest problem. So, it leads to security issues, interoperability issues, integration issues, which leads to increased lead time for solving the problem,” Sivasamban noted further.
This is the typical digital environment in which ThreadSpan steps in and shines, not trying to rip out all the disparate parts, but just trying to sit above everything. “We don’t want to replace each of these tools and we can’t, as all of those are native to their respective technologies.”
But ThreadSpan will sit on top and not only let you see your enterprise IT infrastructure, but also manage it dynamically, with agentic tasks allocated wherever required (with oversight).
Sivasamban offered the most relatable example to explain ThreadSpan’s abilities – propagating policy changes across a large enterprise or organisation. “If there is a policy change that I need to do, security policy change, I’ll do it at one place and it disseminates across all of the regions, countries, devices, etc, seamlessly,” he said.
ThreadSpan is being sold as AI-assisted operations, not AI theatre. Agentic models within ThreadSpan allow tasks to be automated to a large degree, but Sivasamban also acknowledged the trust-building phase, where automation is still controlled to arrive at the desired remediation.
Then Sivasamban connected the dots to the operational KPI that matters most in enterprise: cost per revenue, and speed. For a UK-based logistic firm, which did an acquisition of an entity with 16 odd locations across the globe, with ThreadSpan the firm was able to integrate the entity being acquired – their entire infrastructure – in just a couple of hours. Which otherwise would have been a three-four week exercise, noted Sivasamban.
“I keep telling my Threadspan team that we have to change the definition of MTR, not mean time to repair, but mean time to revenue,” Sivasamban stated. And yes – this is where the AI angle becomes tangible, where AI will help you close tickets faster, deploy policy changes consistently, and integrate new infra without months of pain.
ThreadSpan’s agentic AI capabilities allow CISOs to “talk to their infrastructure platform,” Sivasamban claimed. “If you are the CISO of the enterprise, you can just ask ThreadSpan something like tell me my vulnerabilities of the day, spurring AI agents to sprawl across all of your infrastructure and report back all the vulnerabilities they have seen. And he can further say, you know what, just go fix this SNMP issue. It will go and fix the SNMP issue. So this is nothing but a voice prompt.” This is where ThreadSpan is less “chatbot” and more “ops brain.”
If ThreadSpan was the operational hook, sovereign AI was the geopolitical one, something that wasn’t tacked on as an afterthought. It came up naturally, because data residence and compliance are now part of AI architecture conversations, not just legal footnotes.
In the roundtable, Lakshminarayanan made a blunt point about how the market misuses the term sovereign. “Today, many people are on cloud and public cloud hyperscalers to an extent that is not sovereign. So they might have a cloud instance in India, but not necessarily sovereign.”
That line hits harder in a world where India’s privacy and governance framework is tightening. The DPDP Act and subsequent rules push minimisation, purpose limitation, user control, and breach notifications – exactly the kind of regulatory pressure that makes jurisdiction and governance more than just compliance tick-boxes.
Lakshminarayanan also contrasted Tata Communications with global edge/CDN giants, specifically the likes of Akamai and Cloudflare, which are controlled by the US Cloud Act. It’s a strategic statement, signalling Tata Communications’ India roots, not just an overlay service.
His bigger sovereign AI argument was more nuanced, pointing out how the model might not be sovereign, but the enterprise intelligence lives in the context layer. Which is why India’s sovereign AI efforts need to have an AI operating system.
Lakshmi also acknowledged the uncomfortable reality of how many frontier LLMs are built and trained outside India and how achieving full-stack sovereignty is complicated. But he redirected sovereignty to where enterprise differentiation lives:
“For an enterprise, the context is where the intelligence is and adding that context and bringing that context to a model is where the true benefit will come from. So all of that can be sovereign,” noted Lakshmi.
When it comes to infrastructure sovereignty, he connected it to their cloud and network control. “In Tata Communications, our attempt is to say, if you look at the full stack that we are offering, our networks we manage and operate are sovereign. The cloud we manage and operate is sovereign. The entire AI operating system, if you see how we can bring that entire context layer and how we bring the agentic and we bring the voice, all of this is sovereign. We are the most sovereign player in India when it comes to digital interactions,” highlighted Tata Communications CEO A.S. Lakshminarayanan.
He was careful to define sovereignty as layered – not absolute. “Are we manufacturing chips? We are not sovereign. So we can never be end-to-end sovereign right down to the chip level,” Lakshmi pointed out. That’s the most refreshingly honest framing of sovereign AI. Sovereignty is about what you can practically control – data, context, workflows, network paths, jurisdictional exposure – even if the silicon isn’t yours.
Our conversation ended where it began, on the premise that enterprises are sprinting toward AI without stabilising what’s underneath. “The normal AI itself in the enterprise world, we are only starting to scratch the surface,” Lakshmi noted, rounding up how AI doesn’t just scale enterprise capabilities, but it also scales enterprise fragility.
And when “things happen at machine speed,” the infrastructure will come under tremendous stress, and everyone will have to prepare for the eventuality sooner or later in their AI adoption curve.
Also read: OpenAI flags rising cyber threats as AI models get more powerful
