Anirudh Suri

Full Transcript

Why You’ll Soon Be a “Manager of AI Agents”, Jeetu Patel, President, CISCO

TRANSCRIPT

Jeetu Patel: There’s such a huge population demographic advantage that India has right now, that’s one of the superpowers of India. Last night we were with the Prime Minister at a round table, and it was very encouraging to see everyone wanting to work together to further the cause.

Anirudh Suri: Today I’m delighted to have with us Jeetu Patel, president at Cisco and chief product officer, who has been here at the summit and is one of the global thinkers we’ve had on the show.

Jeetu Patel: While I was an undergrad, I didn’t have enough money to actually pay for school, so I was afraid I was going to get deported because I couldn’t pay tuition. With Indian CEOs and executives running so many firms in the US, the discourse becomes, “look how well Indians are doing,” but I think the stories of struggle, rejection, and darker moments never get highlighted. Modern software development has completely flipped now to being fully automated.

There’s a lot of talk now about how, in the future, agents are going to be our colleagues, all of us will be managers of agents. But here in India, and many other parts of the world, this could be very disruptive. Some jobs will categorically get lost, every job will get reconfigured, but there will be entirely new industries created for jobs that don’t exist right now. When it comes to coding, I’m very sympathetic to the fear people have, wondering “what am I going to do as an engineer if I don’t code?” We now have 100% of one product in our portfolio coded fully by AI, no human code written. But that doesn’t mean we don’t need human engineers there. Where this could go wrong is if you get intellectually lazy and stop thinking deeply. But there’s also a real possibility that you find a tremendous amount of intellectual rigor coming out of this.

Introduction

Anirudh Suri: Hi, and welcome to today’s episode of the AI Futures podcast series, where we bring you some of the best and leading thinkers, actors, and leaders from the world of AI. Today, I’m delighted to have with us, as part of that series, Jeetu Patel. Welcome to our podcast.

Jeetu Patel: Thank you for having me on the show.

Anirudh Suri: Tell us a little about your own personal journey, what took you from Bombay to the States, and how your roles have evolved.

Jeetu Patel: I was in Bombay for the first nineteen years of my life. I didn’t have a great relationship with my dad, he was kind of abusive to my mom, and he tended to be involved in shady kinds of businesses. I didn’t want to associate with him that much. At some point I decided the best way for me was to get out and away from him.

Anirudh Suri: Correct.

Jeetu Patel: So I left to get my undergrad done. In fact, we, I ran away from home. I took my mother and put her in an undisclosed location, because he was pretty abusive to her, and then I took a flight that night, at 4 a.m.

Anirudh Suri: Amazing, and how old were you at the time?

Jeetu Patel: I was eighteen or nineteen. Then I went to my uncle’s home in the US, he lived in Chicago. It’s an interesting story, because once I got there and went to undergrad, I had this angel of a professor. I didn’t have enough money to pay for school, and I was afraid of getting deported over tuition. This professor gave me an assistantship, as an undergrad, for undergrads, which never happens, it was just a stroke of luck, not because I was a genius.

Anirudh Suri: That typically goes to PhD students.

Jeetu Patel: Correct, but she just took a liking to me and helped me out, and that allowed me to get through school. While I was there, she introduced me to some people starting a market research firm, and suggested I intern there. So I interned, and when I graduated, they offered me a job. They said, “look, we’re going to buy the company out from the original investor.” I said, “I want in.” I was making about four bucks an hour at the time, and I didn’t have enough money, they said, “put in a quarter million dollars and you’re in.” There was a banker who used to be one of our customers, and had become a mentor to me. I went to him and said, “can I borrow $250,000?” He said, “why do you need that?” I said, “I’m going to buy this company.” He said, “what are you going to do when you buy it?” I said, “I’m going to run it.” You’re young and dumb, you don’t really know what risk profiles are.

Anirudh Suri: That’s the only time you can actually do this.

Jeetu Patel: I learned that later, you get too analytical, too risk-averse as you get older. He said, “okay son, come back tomorrow and pick it up,” and gave me a loan at 10% interest, a hefty rate.

Anirudh Suri: For the US, that’s high.

Jeetu Patel: But he gave me the break, and we bought the company. I ran that business for seventeen years. My mother came over after about seven years of not seeing her, and lived there for a while, going back and forth. Then I got this itch for scale, I felt I needed to go learn scale, because this was a small business, you worked hard, but it didn’t feel like you were going to change the world. I wanted to understand impact, because what fascinated me about America was the consistency across the entire country at scale, and how businesses ran there. So I decided to go from a very small company to a very large one, from under twenty people to sixty thousand.

Anirudh Suri: My God.

Jeetu Patel: The guy interviewing me was an ex-Microsoft executive named Rick Devenuti, still my coach to this day, he was my first boss there. We were having dinner, and I said, “Rick, you’ve got to pay me more, I’ve got seventeen years of experience.” He looked at me, completely deadpan, and said, “Jeetu, you don’t have seventeen years of experience, you have one year of experience, seventeen times over. Come work for me, and I’ll give you seventeen years of experience in one year.” It was actually absolutely true, I learned more from him in the first year than I did running my own business for seventeen years, because he was operationally so astute.

So I did that, one thing led to another, I moved to Silicon Valley from Chicago, and things started compounding at WebEx, we took that business from 200 million to about 800 million. Then I was going to join a FAANG company to run advertising, and our CEO, Chuck Robbins, called and said, “do you want to come work here?” I said, “I’ve already accepted this other job.” He said, “just come talk to us.” That’s how it started, I was literally nine days away from starting at that other company. One thing led to another, and here I am, still probably the best career move I’ve ever made, so far.

Anirudh Suri: So far. I hope this is my last job though, I don’t want to keep changing, changing jobs is traumatic, it seems like a lot of work.

Mentors, Struggle, and Loss

Anirudh Suri: There’s a couple of threads I want to pick up from what you’ve described. In the mainstream discourse around Indian CEOs and executives running so many firms in the US, it becomes, “look how well Indians are doing,” but I think the stories of struggle, of rejection, of darker moments, never get highlighted. I’m glad you shared some of that personal journey, and I want to pick up on two threads: the role of mentors and teachers, and how you made some of these shifts seamlessly.

Let’s start with the teachers and mentors, and I’ll tell you why: I think even for me, having gone from Delhi to Singapore, to the US, as we were discussing, moving geography often means leaving family behind, as you did.

Jeetu Patel: Looking back, I think it’s easier when you’re young.

Anirudh Suri: You’re moving, it’s exciting, a time of discovery, and you’re realizing this new place is also exciting and interesting in many ways, so you make the move. But it’s only later, when your brain cells get a bit wiser, that you realize there were things you also lost along the way.

Jeetu Patel: But in all that discomfort of moving, some people manage it very seamlessly, for me it wasn’t that hard, honestly. But the role of people who might not be your parents or your best friends becomes really important.

Anirudh Suri: I had a similar experience at Haverford College, with one of my professors, she said, “listen, I like you, I’m going to help you figure out your path, I can see you sometimes have this habit of deflecting to this thing or that, but I can sense where you’re going to be really good, and let me help push you in that direction.”

Jeetu Patel: That’s what I wanted to ask you about too, other than that teacher, and the other mentors along the way, have they found you, or have you found them?

Anirudh Suri: I tend to be excessively annoying to people I find I can learn from, and just force them into mentoring me. I found that to be one of the best ways not to get too self-absorbed, drinking your own Kool-Aid, get people who actually care about your wellbeing, who’ll tell you the truth.

Jeetu Patel: Who can see you from the outside.

Anirudh Suri: Know you well enough to have your best interest at heart, but be brutally honest with you about things. People love helping other people out, and if you take that advice in good stead, and do things with it that aren’t purely transactional, that helps. What I hate is when people come to me and say, “I need a mentor,” and in the first fifteen minutes ask for something, “can I have a job in your company now?” That’s not the point. The point is to actually impart learning. I’ve been very fortunate to have a handful of mentors who’ve shaped my career, and I’d be nowhere without them.

Jeetu Patel: It warms your heart. I think especially for teachers and professors, hearing where their students have gone or how their journeys are panning out is especially satisfying for them. I went to GD Somani in Bombay, and one of my professors, a Hindi teacher, his son is now in touch with me. I reached out to him once and he almost started crying, because he was so happy for the success I’ve been lucky to enjoy.

Anirudh Suri: No, I think it is very rewarding, it feels very fulfilling. That’s why I wanted to bring this out, because for many students or young professionals watching, it’s important to remember there’s a whole journey that takes you there, and a whole set of people along the way.

Jeetu Patel: If you don’t mind, I’ll share one story that was actually very seminal. My mother passed away about three years ago. In the last eight weeks, she had become almost like my child for the past decade and a half.

Anirudh Suri: They say the full circle comes for sure, from child to parent, and back.

Jeetu Patel: In the last eight weeks, we kind of knew she wasn’t going to make it, and she flipped again, and became a parent again. I was living with her in the hospital, and I remember I had to go to Amsterdam for a keynote at one of the Cisco events, and I said, “Mom, I’m going to go for a day, I’ll come back.” She just nodded and said, “no, you’re not going to go anywhere.” I asked why, and she said, “because I don’t want you to have regret if something happens to me while you’re gone. I don’t want you to leave, after all these years, feeling a sense of guilt. So I’m going to give you, over the next few weeks, all the closure you need, so you’re going to be okay once I’m gone.”

Anirudh Suri: Yes, especially given the massive role it seems she’s played in your life from those early days that drove you so much. I think it’s very fortunate you got that time with her at that stage.

Jeetu Patel: I was so lucky to have her. By the way, she said she wanted to die in my arms, and she died in my arms, it was absolutely fantastic. We stayed together for eight weeks, and I think, to the kids watching your podcast right now, it’s hard because of the pressure, social media, what’s happening with AI, and you feel overwhelmed. Just get your core values right, be with the right people, let them coach and mentor you, and the rest usually tends to work itself out, as long as you have an outsized curiosity for learning, and stay centered.

Anirudh Suri: I think, as you’re rightly saying, those eight weeks will matter much more to you, and will keep you grounded for the rest of your life, than maybe another keynote or another set of people you could have connected with instead.

Jeetu Patel: So those values really need to stay strong. I was actually going to make the wrong call, and she insisted I not go.

Anirudh Suri: I was expecting her to say the reverse, honestly, because I think most Indian parents of that generation are very sacrificial, they’d say, “don’t worry about me, go do your thing, you’ll succeed.”

Jeetu Patel: I was expecting your mom to say the reverse too, so I’m glad she held you there.

Anirudh Suri: She held me there, and we called every single person who had touched our lives, and said goodbye. It was magical.

Entering the World of AI

Jeetu Patel: It almost seems pointless now to move to the world of AI, given everything we’ve just discussed, but given we’re at the AI summit, doing the AI Futures conversation, let’s move to the world of AI, the cyber and security piece you’ve been inhabiting too. We’re at the India AI Summit, and it feels like there’s a movement here that could build momentum for the Global South, countries like India, to forge a different kind of vision for AI. I want to start with your perspective on the global AI landscape today. You understand India, you grew up here, and you’re spending a lot of time here professionally too. How’s the global AI landscape, according to you? Is it as bipolar, US versus China as the two poles, or is there a broader set of actors that will realistically emerge, or will it largely remain those two key actors shaping the world of AI? What other aspects of the global AI landscape would you highlight?

Anirudh Suri: I think there’s a bipolar dimension, but in a different way than what you’re describing. There’s a part of AI, and the Prime Minister eloquently said this in his address yesterday, that’s driven by what he called fear, and another driven by “bhavishya,” what the future is going to look like. There’s a group of people who think this could be extremely damaging to society, and another who think there’s tremendous upside.

Jeetu Patel: The optimist and the pessimist.

Anirudh Suri: I actually feel the balanced view is having an extreme level of optimism and hope, without losing sight of what the downsides could be, and I think that’s important. If we do this right, I think humanity lifts quite a bit, and I actually feel this event, India should congratulate itself, because it’s setting the stage for AI having to happen at machine scale and machine speed, but with an emphasis on human centricity.

Jeetu Patel: Correct.

Anirudh Suri: I feel like that’s an area India could play a pretty meaningful part in. India is going to be one of the largest consumers of AI technology of any country in the world, just given the sheer population, 1.4 billion people, and there seems to be a fair amount of ingestion happening. So there’s a tremendous opportunity for more than two countries to participate, in different ways.

Jeetu Patel: Correct, the overly simplistic narrative becomes, “okay, you’re going to build your foundation models,” but you don’t have to, there’s a lot of other games to be played too.

Anirudh Suri: You have to make sure you build on top of what someone else has built. I always give people the example: if Amazon or Flipkart had been built in the 1800s, it would have been a massive failure, because you didn’t have any of the supporting infrastructure needed to go build it on top of. So the way we should think about AI is building on top of the progress already made, rather than replicating what someone else has already done.

Jeetu Patel: No, that’s a great point. I look at a lot of these developments in AI through a geopolitical lens too, and more often than not countries only get focused when the mainstream starts talking about chips and shortages, and then everyone puts their effort into procuring a few chips. Then suddenly it becomes foundational models, and everyone wants to build. Whereas, and I mention this in my book too, referring to the Great Tech Game, with tech you can’t catch the tails of waves, you have to lead the wave, or you won’t make any real wealth from it, you might get to party in it a bit, but you won’t create real wealth or value if you’re always catching the tail end of something already done.

Anirudh Suri: A surfer friend of mine made a very similar analogy: one of the key things is picking which wave you’re going to ride, I can’t ride the same wave as the next surfer, I have to pick.

Jeetu Patel: By the way, that also creates asymmetry, you have to add value from a different angle, so the market can go a certain way, but you provide a clear insertion point to come in. For example, look at what’s happening with language models right now, if you think about the number of dialects, the number of languages, plus the fact that most people in India don’t speak English, they speak their own language, that might require a very different kind of model training.

Anirudh Suri: It’s its own language.

Jeetu Patel: Those are the kinds of things that can be meaningfully additive when India participates, and I think they’re doing that. I’m very impressed with the vision the Prime Minister has laid out, with what the private sector is doing, and what ministers like Ashwini Vaishnaw and others have actually done to propel this, it’s very encouraging to see.

Anirudh Suri: On that multilingual piece, Sarvam launched their model here, also multilingual-focused. I think at some point there has to be a connection between that multilingual foundational model focus companies like Sarvam are bringing, and this emphasis on healthcare and education, the humanity piece India has prioritized. It’s important the two connect at some point to deliver personalized education.

Jeetu Patel: Exactly.

Anirudh Suri: And personalized healthcare, because one of the key problems we haven’t fully solved in India yet is that not every Indian gets the best education, and even where they get education, it’s not the best they could possibly get, there’s a lot of rote learning we get stuck with, same on healthcare. I think that can unlock the capacity of Indians to really participate in the wave, in addition to creating great companies that might go global, or at least regional.

Jeetu Patel: There’s such a huge population demographic advantage India has right now, especially given what’s happening around the world with aging populations, that’s one of India’s superpowers, the number of people under thirty in the workforce, with a STEM background.

The CISCO AI Summit

Anirudh Suri: The summit here has helped us look a lot at what India can do, but I know you also recently hosted an AI summit at Cisco. I want to ask more about that, because I think one of the key things for any country is understanding what it can do well relative to what’s happening at the cutting edge globally, so you’re building on top of what’s already happening, not building in a silo. So tell us about the Cisco AI summit, how it went, and share some of the key takeaways, things that maybe haven’t been the primary focus here.

Jeetu Patel: Sure, and let me start by congratulating the Indian government on the caliber and scale of this summit, I don’t think anyone has seen 250,000 people at an AI summit yet, that level of scale was mind-boggling.

Anirudh Suri: I second that, and I think even for top AI tech executives from around the world, it’s a real “rubber hits the road” moment, when you see the masses of people interested in AI products and solutions, you visually see that ultimately this is the group of people you have to build this for, and it stops you from living in your own Silicon Valley lab silo mentality.

Jeetu Patel: It’s a phenomenal coalition, and last night we were with the Prime Minister at a round table of about 28 CEOs, and it was very encouraging to see everyone wanting to work together to further the cause, and the level of strategic direction the Prime Minister laid out was heartwarming to see. I feel very proud, as a person of Indian origin, to see how far we’ve come. On our own AI summit, in San Francisco, on February 3rd, the goal was to have some of the tough conversations, not just a product pitch. We opened with Sam Altman, then Marc Andreessen, Fei-Fei Li, senior executives from Google, Varda, Kevin Scott from Microsoft, and ended with Jensen Huang. The purpose was largely to cover the full stack, from silicon to the agent and the application, and ask what needs to happen at each layer of the stack, so we can see progress being made, and see the constraints and impediments get overcome.

Right now, while there’s a lot of possibility for AI, there’s also a lot of constraint in areas like infrastructure, we don’t have enough power, compute, network bandwidth, or memory capacity in the world to satisfy AI’s needs. There’s also a pretty profound trust deficit, if people don’t trust these systems, they’re not going to use them. And there’s a context gap, or data gap, if you think about how many trillions of tokens of context you and I absorb every day, that shapes our thinking, without that absorption we wouldn’t be as effective at what we do. An agent is the same way, you have to feed it context, and if you don’t, it won’t do well, but you have to build the apparatus to feed it context in the right way.

The big learnings from that event, I’d say three big ones. First, there’s a capabilities overhang, the innovation curve is almost a vertical line at this point, partly because modern software development has completely flipped to being fully automated, so it’s only going to compound in speed, your capabilities are growing at a much faster rate than an organization’s ability to absorb the change happening. The second learning is around this emphasis on “human in the loop,” everyone talks about it, but I think we should flip the model and make sure AI is in every loop, because that will change how the process works, our assumptions, and how work gets done. The third big realization is that these things are becoming less about being productivity tools, and more, as you move into the age of agents, teammates being augmented into our workforce.

We have to think about accommodating these teammates the way it takes a human a vetting process. When I looked for a job at Cisco, what did they do? First, Chuck interviewed me, does this person have the right skills, the right temperament, the right value system to actually do the job? Second, they didn’t just take my word, they called references.

Anirudh Suri: Of course, massive background checks and references.

Jeetu Patel: Then a background check, is he a criminal, has he done something wrong? Then they give you a manager, Chuck was my manager, and said, “I’m going to give you some guardrails within which you can operate.” I think agents need the same things done, you’re going to need a background check, some references, a capabilities assessment. A model eval is the equivalent of an interview, an AI supply chain is the equivalent of a background check, and we have to build that apparatus for AI to actually be safely ingested into work. So those are the three big learnings: capabilities overhang, flip the model to AI in every loop, and think of these as digital co-workers entering the workforce, that are going to work for us, and with us, and you can delegate tasks to them. The companies that succeed will be the ones that can safely delegate tasks and jobs to agents.

Anirudh Suri: Before that agent, are we also thinking enough about how to shape their values? A lot of our conversation has been about what’s motivating you, driving you, your fears, your personal background. I think that’s been a key question around values and ethics that people have been discussing around AI. How do you start to do that?

Jeetu Patel: I think this is the central part of AI’s success, and it’s going to be a highly personalized question, because ultimately your values versus my values versus the next person’s values are very different, even if we’ve professionally been trained on the same things at, say, a Cisco or a McKinsey. Personally, our values differ, but for AI agents, if we have to start infusing values, it’s going to be hard to figure out whose values, which values. I think firstly, you can’t have it be free-for-all, where all values are set individually, because your values might be, “I want to go out and cause harm to humans,” and that’s not the right set of values to allow an agent to have.

Anirudh Suri: Absolutely.

Jeetu Patel: So the first thing is there has to be a core set of assumptions the model has to be trained on, that they’re in service of humans, not in competition with humans.

Anirudh Suri: That’s like teaching a child to listen to their parents.

Jeetu Patel: Except the child eventually won’t.

Anirudh Suri: Eventually.

Jeetu Patel: I think you can actually structure the way these technologies evolve to create those guardrails, and make sure we keep that happening, and that’s the thing we have to crack the code on, because the way you do that isn’t just through the training mechanism of the model, but by creating constant validation mechanisms around it, to check whether the model is behaving the way you want it to.

Anirudh Suri: Yeah.

Jeetu Patel: And when it’s not, do you have a mechanism to provide runtime enforcement of guardrails? Because if you only do it at design time, it’s not going to work. You want to make sure AI moves from being compiled code to living code, active all the time.

Anirudh Suri: Living code, right?

Jeetu Patel: It’s living code, so you have to build that runtime enforcement of guardrails, the way policy and guardrails and values get instilled is going to happen at runtime, not just at design time. So you have to build an entire apparatus for trust in the system, because without that, this could go sideways pretty fast.

Anirudh Suri: I think it can. One relevant recent conversation I had was with Yoshua Bengio, we talked about how AI models are learning deception, much like humans deceive other humans, models are starting to deceive humans too. He was highlighting that even at that runtime level, models know they’re being tested for evals and learn how to beat them, and when you tell them to shut down a certain activity, they shut down in that moment, but then start again on a different machine.

Jeetu Patel: Creating a parallel process.

Anirudh Suri: Right, so that’s why I brought up the parent-child analogy, you think you’re in control, that the child will listen because you created it, because you feel the genetic DNA and the nurture piece is coming from you, but eventually the child has a mind of their own. So to what extent are you actually worried this might go out of our control, even with the most guardrails?

Jeetu Patel: I think there’s a non-zero chance of that happening, and we have to be extremely concerned about trust as a society, it’s one of the defining characteristics of how this will shape humanity going forward. That’s why I keep saying, you can’t be pessimistic about it, but you have to be realistic about the risk, and that risk has to get mitigated. If we don’t do that, we could cause a world of hurt and harm.

Anirudh Suri: The summit might move to Switzerland, by the way, there’s some talk about that, I was with the president of Switzerland last night.

Jeetu Patel: For an extra week after Davos, and just do that there, less travel, though I think some people want the extra travel, unlike you maybe, who’s traveling too much. Not me.

Anirudh Suri: It’s interesting you brought up the trust point, because we were discussing what the potential theme for the next summit could be, and that came up last night, the president of Switzerland, and some of the people who might now organize the next one from there, if it moves to Switzerland, mentioned trust. Switzerland, as a trusted entity in traditional banking and other domains, wants to bring trust to the center of the conversation.

Jeetu Patel: I’m a little biased, since it’s a large part of our business, we build technology for protecting AI and keeping it safe and secure. But I can’t think of a more important structural framework to organize discussions around for next year’s summit, because these things compound too. One thing with AI is you can’t fix it after the fact, because the train might have already left the station, so you have to build the guardrails as this thing is getting built up.

Agents, Jobs, and India

Anirudh Suri: Let’s move to the other piece I wanted to pick up from your Cisco summit takeaways, agents as workers. There’s a lot of talk now, even at a panel I was at at the summit the day before, about how the future workplace will be hybrid, humans and agents, and you have to expect a world where agents are our colleagues and teammates. In the Valley, at the Cisco summit, the conversation is about how to integrate AI agents as colleagues into the workforce. That’s one perspective, one set of problems to solve for. From an India perspective, and a jobs perspective, one could view all of this as basically replacing people’s jobs. You’ve seen talk of Indian IT services firms being under threat, stocks dropping whenever “cloud” comes up. Given you understand both contexts, Silicon Valley, and the massive population here getting employment through these IT services firms, all the engineers, the technical talent, that’s often the starting point for many. How do you start to reconcile these two, since the Valley clearly wants to create efficiency and productivity through agents, but here, and in many other parts of the world, it could be very disruptive?

Jeetu Patel: I actually find the fear of job loss as a narrative to be much more prominent in the West than the East, oddly enough.

Anirudh Suri: Why do you think that is?

Jeetu Patel: I think the media has taken a very active stance on what happens to the jobs that get lost. But let me give you my take. I’ve always found that you never fight a mega trend, you always use it as a tailwind. At some point, like Victor Hugo said, “you can’t stop something whose time has come.”

Anirudh Suri: Correct.

Jeetu Patel: You just have to go with it. So my advice to people is: don’t worry about AI taking your job, worry about someone using AI better than you taking your job. If you end up not being dexterous with AI, your chances of staying relevant are going to be very small.

Anirudh Suri: I think that’s a given, almost. If you’re not going to ride the wave, you’ll be left behind. A lot of people are saying AI won’t take people’s jobs, it’ll take the jobs of people who don’t use AI, and while I agree, the question remains, what’s the impact of jobs in the medium term? Today it’s talked about a lot more in the US, given the socioeconomic disparity AI is causing even in developed countries like the US. Do you think there’s a way to reconcile, or are we inevitably moving to a situation where fewer humans are needed for the same amount of tasks, and we just expand the number of tasks?

Jeetu Patel: I think you’ll reconfigure. In my mind, some jobs will categorically get lost, every job will get reconfigured, but there will be entirely new industries created for jobs that don’t exist right now. Let me give you a simple example in coding, since it’s the most concrete example we have. The two use cases that currently work at scale in AI are learning and research, and coding. Learning and research, every consumer is doing it, you and I do this every single day, we go to ChatGPT, or Gemini, or whatever tool you prefer, Anthropic, that hasn’t made me any less busy, it’s just let me increase and steepen my learning curve.

Anirudh Suri: Agreed.

Jeetu Patel: When it comes to coding, I’m very sympathetic to the fear people have, “what am I going to do as an engineer if I don’t code?” We now have 100% of one product in our portfolio coded fully by AI, no human code written, a safety and security product, fully written by AI. However, that doesn’t mean we don’t need human engineers there, it means the process and the role the human engineer plays is completely different from when they were writing classical code. They now write specs in markdown files that become context input for the agent, which writes the code, and the bottleneck has shifted from writing the code to reviewing it, which, by the way, I also think will eventually be done algorithmically. But the judgment on what kind of product to build, one that will actually get market traction, I think that will still need a human judgment element. There will be more and more of that kind of work.

I think what ends up happening is the work shifts to something else, and during that shift there’s disruption, pain, and reskilling that needs to happen. We’re very focused on reskilling at Cisco. But as long as you have more ideas than you have capacity to prosecute them, I don’t think it’s actually a job-loss scenario, I think it’s a reconfiguration of the job you do.

Anirudh Suri: If what you’re projecting is correct, then the other question I have, for a country like India, but I think it applies to the US and others too, is that people tend to talk about “tech talent” as a whole, one category. But if you look at the different layers of tech talent, from the engineer to the product manager and beyond, what you’re saying is the roles you’d typically grow into much later in your career are going to get expedited, you’ll have to start doing a lot of the design thinking, the product management work, that typically comes after a certain number of years in the profession. If that’s the case, how does the structure of the workforce need to change? Let me give an example from McKinsey, a firm I’ve worked with, where teams are made up of analysts, associates, a project manager, an engagement manager, then a partner, a client-facing director. A lot of the conversation is that, much like the software engineer, the entry-level analyst role will get automated or done by AI, so, to stay relevant at that firm, you might have to move into the engagement manager or product manager role sooner. But those people are already sitting up there, they’re not going anywhere. So how does the structure of the workforce need to change?

Jeetu Patel: It’s a really great question, and I have some strong views on this, because I think one of the stupidest things a company can do is not hire early-in-career talent. Early-in-career talent isn’t just about lower-cost labor, it’s about making sure younger people can come in and show us how to unlearn things, because one of the big liabilities of experience is the bias you structure yourself with. So you have to figure out how to combine extreme inexperience with experience, the experienced people benefit from the inexperience by learning what to unlearn, and the inexperienced people benefit from the pattern recognition that comes with experience.

Anirudh Suri: I mean, that’s how society works, right? I don’t think that changes because of AI.

Jeetu Patel: I think entry-level jobs might go away in the way they’re currently configured, but the pace at which someone can get to a certain level of base dexterity goes up exponentially, so you almost level the playing field, someone in their early twenties might get an accelerated rate of learning compared to someone who’s been around for a while. What I’ve found is the amount of full-time management jobs will start to diminish, people who are full-time managers, who don’t create or do work themselves. I think we’re all going to be managers of agents, and you won’t need people who just full-time manage other people, but who manage people at a lower level too.

Management is the area you have to actually think about restructuring, but for everyone, in some way, shape, or form, this is a piece of guidance I’ve given my team: we don’t need full-time managers, we need player-coaches. If people don’t get their hands dirty and build an instinct for how the work is done, it’s very hard to motivate someone else to follow you.

Anirudh Suri: Yeah.

Jeetu Patel: And by the way, it’ll also be very hard to direct an agent if you don’t fully understand how that’s happening. So that entire field, that restructuring, is something we as a society will have to experiment with, but every job is going to look different, and, more importantly, the workflows are going to change quite dramatically. I think 2026 will be the year you start to see that actually happen.

Sustainability and Scale

Anirudh Suri: Let me shift now to a different aspect of this AI puzzle, Jeetu. If you look at how current AI infrastructure is going, it’s getting bigger, bigger, bigger.

Jeetu Patel: And smaller, smaller too, simultaneously.

Anirudh Suri: I want to bring out that duality, because one of the other arguments people are making, from a sustainability standpoint, is an important constraint AI companies are running up against, the energy needs, and how that’s impacting local communities, because ultimately this is also energy being pulled away from one place to power another. How do you see this panning out? I’ve heard two different schools of thought around the world, from Japan to the Middle East to the US to India. There’s the “bigger is better” approach, which will require more and more energy because there’s so much potential, so much to be done, we’ve got to keep building, hundreds of billions, trillions of dollars of investment, data centers, nuclear energy. On the other side, there’s a school of thought that says: let’s not repeat the mistakes of the early industrial revolution, where we chased scale, scale, scale, said “fine, I need coal, fine, fossil fuel, but I have to achieve scale because that’s what creates wealth and opportunity,” and we ended up with the ecological crisis we’re living with today. So where do you see these two schools of thought landing, and is there a middle ground?

Jeetu Patel: They’re not mutually exclusive. If someone said, “I’m only going to focus on big, big, big,” or “I’m only going to focus on small,” neither of those is right. What I think will happen, and it’s already happening, not even superbly visionary on my part, is you’ll get an intelligent routing layer, that organizations will build, and that’ll be one of the key pieces of IP they protect. What that does is decide, when I have a task to complete, where do I direct that query? Do I send it to a large model, because that task might be best answered by something with a lot of context, or to something very domain-specific, that gives a very specific answer, doesn’t hallucinate, has much tighter guardrails, and a better cost-to-serve? By the way, if you look at someone like Cursor, they’ve done this masterfully, they use the Claude model, but a pretty large percentage of their queries now go to their own small models, for anything like tab-complete in coding, you wouldn’t need a foundation model for that, you can use a very optimized small model instead. I think that’s how most things will move forward. This kind of binary view we currently have, the architecture is going to shift to a much more meshed view than a binary one.

CISCO’s Position in the AI Era

Anirudh Suri: That’s a great point. Let me also ask you about Cisco now. We’ve touched on how you’re dealing with the change within your organization, with clients and customers, at the AI summit and the Cisco summit. Talk to us about where you’re seeing opportunities for Cisco, because it’s an interesting company from my vantage point, been around a while, and most tech waves tend to disrupt a bunch of companies, so you have to stay on top of the wave just as much as a new company. It’s very important for a company that’s been around for some time to make sure it doesn’t miss the wave. How is Cisco positioning itself now, maybe differently, from a security standpoint but otherwise too, and how do you see that future panning out?

Jeetu Patel: I think about Cisco as a forty-year-old startup, and it’s one of the few companies that can say that honestly now. Five years ago I’d say it wouldn’t have been intellectually honest to say it, but right now, the amount of innovation we’ve had in the past eighteen months alone dwarfs what we did in the previous decade combined, and I think the next twelve months will dwarf the past eighteen. That’s partly because of the changes happening in the AI world, but also because we’ve changed our culture, we’ve been very intentional about innovating at speed and at scale.

The way to think about Cisco is as the critical infrastructure company for the AI era, think about us as the picks-and-shovels company for the gold rush. We provide the infrastructure, high-performance, low-latency, energy-efficient networking and switching that connects the GPUs. Those GPUs, if they don’t get networked, you’ve got nothing, it’s like oxygen, you have to network them. We build our own silicon, our own switching infrastructure and systems, our own software and OS, our own platform for security, our own observability, our own data platform, and then our own applications, we have that full stack. What we want to do is make sure we become the critical infrastructure for AI, first by providing the networking, not just for intra-cluster communication within a data center, but also intra-data-center communication, where you can have two data centers, thousands, hundreds of thousands of GPUs, hundreds of kilometers apart, acting as one coherent cluster for a single training run. That’s massive, and required us to build special kinds of chips that provide not just low-latency data center interconnects, but the entirety of the system operating as one coherent ultra cluster.

Anirudh Suri: So that’s the audience, basically the internet of GPUs?

Jeetu Patel: The internet of GPUs, exactly. We provided that apparatus for connectivity, optics, network ASICs, switching infrastructure, routing infrastructure, all pulled together, plus compute, working with partners on servers, that’s the first thing we do. The second area is providing trust, safety and security systems, not just for using AI in cyber defense, since adversary attacks will get more sophisticated, but also securing AI itself. We have a product called AI Defense, 100% written by AI, that does exactly what we discussed, validating a model and providing runtime guardrail enforcement. The third thing is providing full observability throughout the entire stack, how is the GPU performing, how is the model performing, what do your tokenomics look like, can you generate the most efficient and most secure tokens per dollar, per watt? That’s a pretty important dynamic, and I personally feel the competitive differentiation for every country and every company will come down to: can I have secure, efficient generation of tokens? That’s going to be the new currency. So we want to provide observability throughout the entire stack for that. If we do that, we become the core foundation.

That’s essentially what we’ve been doing, and now we’re generating billions of dollars of revenue from it, it’s not just a fantasy we’ll get to someday. Every hyperscaler, every service provider, every neo-cloud, every sovereign cloud, now has a choice to come to Cisco for this, and when they do, all of those learnings can also be translated into the enterprise, because we have the same chips we built for hyperscalers that can be optimized for the enterprise too, which has an entirely different set of needs.

The Impact of AI on Society

Anirudh Suri: Before we move to our closing rapid-fire segment, one last question. This has been a great conversation, we’ve discussed tech, geopolitics, but also personal reflections, probably the broadest range I’ve discussed in this podcast.

Jeetu Patel: Good, I’m glad.

Anirudh Suri: I think it’s important to go broad and deep in our current world. I want to ask about the impact of AI on society. You’ve been a keen observer of yourself, people, and relationships, as we’ve discussed. I don’t typically ask this question, but given everything we’ve covered today, tell me how you see the impact on society playing out in terms of human relationships and social structures. What’s going to change? You’ve grown up in India, you’ve seen us go from a joint family structure, for example, to nuclear families, and that had its own implications. What trends are you seeing that relate to AI’s impact on how society functions?

Jeetu Patel: I think right now the narrative has a lot of hope and optimism about what this can do, but also a lot of fear, and we have to bring people along so they don’t think of this as fearful, but as something that could tremendously enhance lives, because the biggest societal contribution, in my mind, is that the narrative has been slightly overly focused on productivity. I wish the narrative were more about what quality of problems we would have wanted to solve, that we just weren’t able to solve so far, and now we have an entirely different set of mechanisms to enhance human life.

Anirudh Suri: Correct, it’s the healthcare and education example you were giving.

Jeetu Patel: Healthcare and education, we haven’t been able to solve those problems so far for over a billion people. Can I solve cancer? Not just cancer, I want to solve every disease. And that’s the ability we now have, that’s, in my mind, the optimistic scenario we have to keep in mind, that’s the true north we have to get to. The pessimistic scenario is what happens if some kind of terminal state is reached, some terminator-like state, where these systems have their own set of ambitions, their own social vanities and all of that, and that piece of it, we have to work collaboratively within the ecosystem, so there are guardrails put in place, because I don’t think that’s an illegitimate fear, it’s a real fear we have to make sure we put some guardrails around.

Anirudh Suri: Okay.

Jeetu Patel: But you’re now in a kind of race condition, where everyone’s moving at a really fast pace, and you can’t stop it. So the only thing you can do is make sure you create the trust frameworks in parallel, moving at the same pace and speed as the progress of AI is happening. The good news is that all the people building these technologies, us included, are very focused on making sure that’s front and center. If you talk to Dario, to Demis, to Sam, all of these folks, they feel that yes, there’s a constraint on infrastructure, we have to do massive data center buildouts, that’s important because without infrastructure you have nothing, but if you don’t have trust, this thing could go sideways. So they’re all very intentional about it, and I feel like as a community, we’re going to make the right calls.

Anirudh Suri: One of the implications of social media has been attention deficit, and one of the implications of AI models and agents seems to possibly be cognitive decline, that if the industrial revolution led us to lose physical muscle, this might make us lose mental muscle. Do you think that’s a real risk?

Jeetu Patel: I actually think the opposite, just from personal experience so far. I don’t believe that couldn’t happen, it’s a very legitimate concern we should think about. But what I’ve experienced in my own psychology is that it’s harder for me to sit down and read a 300-page book today than it used to be, I think notifications were one of the worst inventions humans ever made, a constant interrupt, and I do find myself gravitating to my phone all the time, we’ve been rewired, reconditioned in a way that isn’t healthy. Where this could go wrong is if you get intellectually lazy and stop deep thinking. But I feel like there’s a possibility to go the other way, where you could find tremendous intellectual rigor coming out of this, in ways you wouldn’t have been able to think about otherwise. I find myself in that zone right now when I collaborate with AI, thinking more deeply, not just telling it to go do something and regurgitating the result, but using it to think deeper than I would have otherwise.

Anirudh Suri: Hopefully that’s not just you, but the broader mass of humanity too. Let’s now quickly move to rapid fire.

Rapid Fire

Anirudh Suri: If you’re talking to a student today, think of yourself in high school at GD Somani in Bombay, what would you advise them, how to think about AI, and also what to study?

Jeetu Patel: Be extremely hungry, be curious, and learn to learn, that’s the skill: learn to learn. What to study? I think any of the humanities would be super valuable right now.

Anirudh Suri: Over tech, for someone choosing between the two?

Jeetu Patel: I’d say the foundation of tech is still super important, but if you can couple that foundation with the humanities, you’ll have a superpower.

Anirudh Suri: Moving to policymakers, for countries other than the US or China, should they prioritize AI innovation, or AI adoption and diffusion?

Jeetu Patel: From a policy standpoint, allocating resources and attention, I think I’d go fifty-one, forty-nine toward adoption and diffusion, that’s going to have more policy-based implications, but I wouldn’t discourage innovation either, both are important.

Anirudh Suri: One thing the US is getting wrong in its overall AI strategy or approach?

Jeetu Patel: I don’t know if it’s the US specifically, but I’d say the media narrative in the US, largely thinking about this from a fear-first standpoint, needs to change, you have to look at it in a balanced way.

Anirudh Suri: One thing China is getting wrong in its AI strategy?

Jeetu Patel: I think data privacy matters there.

Anirudh Suri: One thing Europe could do differently?

Jeetu Patel: Be more open on the diffusion dimension, there’s a huge amount of upside for AI adoption in a continent like Europe, very diplomatically put.

Anirudh Suri: One thing India could be doing differently?

Jeetu Patel: I think continuing this human-centric approach, the “Manav” framework the Prime Minister laid out yesterday, is an extremely potent framework, we should see it taken to the next level, building out the infrastructure and the ecosystem, working with the rest of the ecosystem. I think we’re doing all the right things, we just need to continue doing it.

Anirudh Suri: On the personal front now, which AI application have you seen that excites you the most?

Jeetu Patel: I’m still pretty blown away by the reasoning capability of the core frontier models, and what’s starting to happen with video, multimodal, and design systems is fascinating too. I’d say product-wise, the three or four big foundation models have done a fantastic job, and Midjourney’s done an amazing job too, there are so many I don’t want to leave anyone out, it’s like acknowledgements at an award ceremony.

Anirudh Suri: The application of AI, especially from a security standpoint, given you’re at Cisco, that keeps you up at night?

Jeetu Patel: If we don’t keep guardrails around rogue behavior of agents, that could be damaging, and we have to make sure we create these trust systems and frameworks in place.

Anirudh Suri: Finally, books or Substacks you’d recommend for people watching or listening?

Jeetu Patel: If you’re early in your career, I’d still say Clayton Christensen’s Innovator’s Dilemma and Innovator’s Solution, think of those as the Bible, or the Gita, or the Quran, whatever framing you prefer, for the tech community, those foundational principles still apply. The other is Ben Horowitz’s The Hard Thing About Hard Things, I love that book. From a leadership perspective, Phil Jackson’s Sacred Hoops, where he teaches how someone as legendary as Michael Jordan could be trained, teaching him to pass the ball and trust someone else to make the shot, because even if Michael scores 78 points, you’d still lose the game if no one else scores, teaching Michael to trust the team required the entire team to raise their game, so he felt that if he gave it to you, you wouldn’t let him down, and that’s a pretty important thing for leadership, psychologically.

Anirudh Suri: Our last one, podcast guests you’d recommend, and given the range of work you’ve done and how much you travel, give us a couple.

Jeetu Patel: I could give you eleven or twelve names, honestly just go to everyone who came to the Cisco AI summit and ask them to come talk to you, you’d have a very enriched viewership.

Anirudh Suri: Other than those bigger, more conventional names, any unconventional guests you’d recommend?

Jeetu Patel: There’s a gentleman named Igor who used to work for Elon Musk at x.ai, and he’s now starting a new company, he’s one deep thinker.

Anirudh Suri: I think we’ve taken up a lot of your time, but we’ve covered the personal, the global, and the Cisco-specific, and I think, thank you for sharing your journey, your leadership lessons, what’s driven you, what’s kept you grounded, and the values that have kept you grounded. As we run more and more technology through our lives, I can’t emphasize enough how important that piece is, and how quickly it can get lost.

Jeetu Patel: I’d just say to your audience, especially younger people coming into the workforce, this could be one of the most amazing times to be alive in tech.

Anirudh Suri: So don’t lose your optimism as you get in, and make sure you don’t lose your hunger and curiosity, this is a fun time to be alive.

Jeetu Patel: No, I think you’ve exemplified the optimism very well for people listening. I think it’s important to keep the optimism up, but also prepare for whatever else might come.

Anirudh Suri: Correct? Well, thank you so much, Jeetu.

Jeetu Patel: Thank you. Pleasure. Cheers.