Welcome back to the AI update. It is really good to be here. Indian IT giant Infosys has officially partnered with Anthropic to deploy AI agents across their enterprise network. Which is honestly a massive signal hiding in plain sight. You see these corporate handshakes all the time and you might just brush it off as another chatbot integration. But that completely misses the point of what is actually happening. We were looking at a fundamental rewiring of a $283 billion IT industry. So what happens when the business of outsourcing human labor gets replaced by orchestrating software? That is the exact tension we are seeing in the market right now. We are going to look at the new Anthropic Center of Excellence and exactly how these legacy systems are being transformed. And we will get right into that after this very short break. To really understand this shift, you have to look at how this industry has operated for decades. Right. The traditional model is incredibly straightforward. Yes. For a long time, that $283 billion industry has run on a very specific fuel, and that is human labor. Bodies in seats. Exactly. You hire smart people, often in lower-cost geographies like India, to write the code or manage the servers or answer support tickets. It is the classic outsourcing model. But what Infosys is announcing here with Anthropic is a hard pivot away from that. We are seeing a move from outsourcing humans to orchestrating agents. Orchestrating agents, which sounds very futuristic, but also a little ominous if you are one of the humans used to being outsourced. Well, it is the existential tension of the decade for the services sector. So here is the key point. We are pulling this discussion from the transcripts of Emphasis' Investor Day 2026, which just wrapped up. Right. Along with their official partnership announcements from February 18th, 2026. and the press release is very specific about the wording they use. Very specific. They are not just deploying generative AI. They are deploying AI agents. Why did they choose that specific word? It is probably the most important word in the entire document. A chat bot, like the early versions of chat GPT or Claude, is passive. It is conversational. You ask a question and it gives an answer. Exactly. It just sits there waiting for you. But an agent represents action. An agent does not just tell you how to reset a password. It actually does it. Right. It logs into the system. It navigates the user interface. It resets the password and then emails the user. So a chatbot is basically a consultant offering advice, but an agent is an actual employee doing the job. That is the perfect way to look at it. Yeah. And Infosys is betting that their enterprise clients, the massive banks and telecom giants, they do not need more consultants giving advice. They need digital employees to do the grunt work. Yes. And they have a very clear roadmap for this. They are not just releasing wild agents into the corporate network to do whatever they want. Which would be a total disaster. Absolute chaos. So they are rolling this out sector by sector, starting with telecom. Okay. And once they prove the agents can handle that piece, they move to financial services. Then manufacturing. And finally, software development itself. Starting with telecom is an interesting choice. Why not finance, which seems to have the deepest pockets? Telecom is high volume and high complexity, but the data is extremely structured. It is basically the perfect testing ground. Because you have millions of logs and signal data and routing tickets. Right. And if an agent messes up a network configuration, it is bad, but you can usually fix it in the software layer. Whereas if an agent messes up a swift wire transfer in financial services. That is a resume generating event for everyone involved. You do not want to accidentally wire a billion dollars to the wrong account. So they start where it is safer. Which brings us to the partner choice. Why Anthropic? Yeah, they could have gone with OpenAI or Google or even open source llama models. But they went all in on Claude I assume this links back to the constitutional AI and safety branding Anthropic is known for It is the primary driver here Infosys brings the domain depth They know exactly how a bank messy legacy backend works They know where the skeletons are in the closet. Exactly. But Anthropic brings frontier models with a safety-first design. In a highly regulated industry, you cannot have an AI that hallucinates. You cannot have a model that accidentally leaks customer data into its training set. And Anthropix architecture is built specifically to prevent that. This follows their push into the enterprise sector last month. Right. In January 2026, they launched Claude Cowork. Which was their direct play for the white-collar desktop, helping individual workers with office tasks. So if Claude Cowork is the individual tool for a worker, this Infosys partnership scales that up to the enterprise level. Exactly. Anthropix builds the engine and Infosys builds the chassis. The transmission and the steering wheel. And they are building it inside something they are calling the Anthropic Center of Excellence. The COE. Vivek Sinha from Infosys called this a milestone moment. But every massive company has a center of excellence for something. Usually it is just a conference room with a nice plaque. And slightly better coffee. Right. Is this one actually doing anything different? I was deeply skeptical too until I looked at the engineering capabilities they listed. They're not just doing basic prompt engineering. They're doing real technical work. Deep technical work. The first one that stood out was knowledge distillation. Okay, so let us look closely at that. Knowledge distillation. To me, that sounds like making moonshine in the server room. Not quite, though. The concept of boiling something down to its essence is very similar. Think of the massive frontier models like Claude 3.5. They are incredibly smart. They know French poetry and quantum physics? Right. They are like a brilliant PhD student who has read every book in the entire library. But they are massive and very expensive to run. Extremely expensive and computationally heavy. Now, if you are a bank loader, you do not need a PhD student to process mortgage applications. You need a smart clerk who knows mortgage rules perfectly and nothing else. Exactly. Knowledge distillation is the process where the massive teacher model teaches a smaller student model only what it needs to know for a specific task. So you strip away the poetry and the quantum physics. And you are left with a model that is incredibly good at checking credit scores, but literally nothing else. And crucially, that smaller model is 100 times cheaper to run. And 10 times faster. Margins are everything in enterprise tech. Because if every single query costs a dollar, you go broke very fast. But if it costs a fraction of a cent, you have a sustainable business. If Infosys can sell a service that costs them pennies to run because they distilled the model. But they charge the client based on the value of the outcome. That is where the profit lives. That is how they survive this industry transition. The next capability on their list was large-scale agent engineering and orchestration. We touched on agents earlier, but orchestration sounds like a massive headache. It is arguably the biggest engineering challenge in AI today. Building one agent to check your email is easy. But what happens when you have 5,000 agents running simultaneously inside a global supply chain? Absolute chaos. Imagine 5,000 interns running around a corporate office without a manager. You need a manager AI. Yes, an orchestrator. An AI that assigns tasks and resolves conflicts. So if two agents try to update the exact same database record at the same time, who wins? The orchestrator decides. Infosys is claiming their center of excellence has actually solved this orchestration layer. So it is like being the conductor of a very robotic orchestra. That is a very apt analogy. Without the conductor, it is just noise. Which leads right into the third point, system intelligence for complex legacy and hybrid environments. The polite term is legacy. The real term is spaghetti code from 1995. We are talking about mainframes and green screens and COBOL code that no one has touched since the Y2K scare Exactly And you cannot just plug a modern API into a mainframe It does not have one It was built before the internet even existed So that is the billion problem How does a shiny new AI agent interact with that? Do they just rewrite the old code? Never. That is way too risky. Usually it involves building complex wrappers or connectors. Infosys says they're using AI to build the bridge. So an AI that can read the old green screen and understand what is happening. And then translate that into modern data for the Anthropic agent to use. It is literally like teaching a new robot to use an old metal filing cabinet. That makes sense because banks are certainly not going to rip out their core mainframes just to use a new chatbot. The risk is too high. You have to meet the legacy tech exactly where it is. If Infosys can solve that translation layer seamlessly, that alone justifies the partnership. There's one more technical capability here. Synthetic enterprise data for safe training. Yes. This one feels slightly counterintuitive. We always hear that data is the new oil. Why would you want fake oil? Because the real oil is toxic. Or rather, it is legally radioactive. You mean privacy laws like GDPR. Exactly. Say you want to train an AI to detect fraud in credit card transactions. You have petabytes of real transaction data. But you cannot feed that into a cloud-based model. No, because it contains real names and real credit card numbers and real spending habits. That is a massive data breach waiting to happen. Right. So you use the center of excellence to generate synthetic data. The AI analyzes the statistical properties of the real data. The patterns and the variance and the frequency of the fraud. And it creates a completely new data set that looks mathematically identical, but contains zero real people. So it creates a John Doe who lives on 123 fake street, but spends money exactly like a real banking customer. Precisely. You can train the model on that safely and crash test it without any risk. And then deploy it on the real data later. It bridges the deployment gap. It allows innovation to happen without the compliance officer having a heart attack. And that leads us to the governance stuff they listed. Cost optimization and reliability and observability. Which sounds boring to us, but for a CIO, observability is how they sleep at night. They need to know why the AI made a specific decision. If an agent denies a business loan, you cannot just tell the regulators that the black box said no. We need a trace. And Infosys is building that governance layer. So here's what happened with the financials. Infosys released some stats during this investor day. The numbers are very revealing, but you do have to read between the lines. They announced they are currently working on 4,600 AI projects. 4,600 sounds like a massive volume. It is a high volume, but project is a very vague term in consulting. Right. A two-week proof of concept is a project. And a five-year global transformation is also a project. We do not know the exact mix. However, they did reveal that AI services accounted for 5.5% of their total revenue in the December quarter. And CEO Salil Parikh called that robust. So what does that mean in plain terms? Is 5.5% actually robust? It is decent. You have to remember that Infosys was the very last of the big three Indian IT firms to actually break out this number. They were playing it very close to the vest. Exactly. So how does that compare to the industry leader, TCS, Tata Consultancy Services? TCS recently released their Q3 results. They reported $1.8 billion in annualized AI revenue. Which, if you do the math, comes out to about 5.8% of their fiscal year revenue. So Infosys is at 5.5% and TCS is at 5.8%. They're basically neck and neck. That is surprising because the narrative in the market has been that TCS is way ahead and Infosys is scrambling to catch up. That narrative was entirely based on silence. Because Infosys was not reporting the numbers, analysts just assumed they were weak. But this data shows they are right in the mix. The main difference is strategy. TCS has been very quiet and building internally, focusing on their own Wisdom Next platform. While Infosys is taking a much louder partnership approach Partnering with NVIDIA earlier and now this major push with Anthropic they are making noise to show they are on the cutting edge It feels like Infosys is trying to win the mindshare battle even if the actual revenue battle is currently a tie. They absolutely have to. And this brings us to the existential pivot. Right. We have talked about the tech and the horse race between these massive companies. But there is a much bigger threat looming over this entire sector. We have that Raiders report regarding the broader Indian IT landscape. It highlighted rising investor concerns. Yes, and the core concern is simple. The traditional business model is about to break. Broken how, exactly? These companies are incredibly profitable giants. For now, they are. But their model is based on labor arbitrage. The old time and materials trick. Exactly. I hire a developer in Bangalore for $30 an hour, and I bill them to a client in New York for $100 an hour. The margin is entirely on the human hour. The more hours a project takes, the more money Infosys makes. But what happens when an anthropic agent running on Infosys Topaz can write that exact same code in three seconds for a cost of five cents? You cannot bill three seconds of work to a client. Right. If efficiency goes up by a thousand percent, your billable hours go down by 90 percent. If Infosys and TCS just stick to the old model AI, it's going to destroy their revenue stream. They are effectively designing the technology that eats their own lunch. That is the ultimate innovator's dilemma. So this pivot to agentic AI is not just about offering a cool new service. It is about fundamentally changing how they get paid. It has to be. They're trying to move from selling effort to selling outcomes. Selling the hours worked versus selling the result. Exactly. Instead of charging you for the 50 people managing your call center, they charge you for 10,000 resolved customer tickets. Regardless of whether a human or a robotic agent solved them. Right. If they could successfully make that shift, AI suddenly becomes a massive margin booster. But if they cannot make that shift and clients still insist on paying for hours, their revenue simply collapses. So Infosys Topaz and this new center of excellence, these are essentially the vehicles for this transformation. They are the lifeboats to get them from the service economy to the outcome economy. The lifeboat implies a rescue. I would call it a rocket ship. They want to be the ones managing the robots. They are betting that the sheer complexity of these agents is so high that clients will still need a massive partner to manage it all. The orchestration and the safety and that legacy integration we talked about. So the work does not actually go away. It just changes from writing code to managing the thing that writes code. Correct. But that requires a completely different skill set and a completely different sales pitch. It is a really bold bet. But looking at the speed of Anthropix development, it seems like the only bet they can make. It is disrupt yourself or be disrupted. There is no third option here. So here's what I want you to hold on to from all of this. The technology has officially moved from chat to act. Agents are here to do actual work, not just talk. This is operational AI. Second, the barrier to entry is deep engineering. You cannot just plug these models in. You need knowledge distillation and synthetic data and deep legacy integration. That is the moat Infosys is trying to dig. And finally, the scorecard is much tighter than we thought. Infosys and TCS are in a dead heat. But the finish line keeps moving. The winner will not be the company with the most engineers. It will be the one with the best agents. It really changes your perspective on what an IT company actually is. We have spent 50 years building an industry around outsourcing tasks to people. And the next 50 years will be about outsourcing tasks to intelligence. So that leaves one question for you to consider. If these IT giants are successful and they actually build autonomous agents that can handle complex tasks end to end, does the outsourcing industry eventually just become the software management industry? At what point does the human element of IT services disappear entirely? The answer to that might come a lot sooner than we think. Thanks for joining us for this update. Catch you next time.
