Cisco President:
What Makes Agentic AI Transformation Work?

Jeetu Patel: We're way past pilot at this point. We are way into production. By the end of next year, the goal that we had was 70% of our products will be 100% written with AI.

Every workflow gets rethought with agents

Michael Krigsman: AI will soon write much of the code for Cisco's products. Jeetu Patel is the company's president and chief product officer. Jeetu, agentic transformation is new. What is it?

Jeetu Patel: We are kind of going into the second major phase of AI. The first phase was when we had this kind of magical experience when ChatGPT came out. You ask a question, you get this answer, and it was this amazing way to go back and forth and get natural language to be the interface that you communicate with computers and machines. We're now in the second phase with agents. And frankly, I thought the second phase would have come a little earlier with OpenClaw.

But where I think the industry had underestimated the importance of was just the frictionlessness of how people can start using agents. So not everyone wants to go out and have brain surgery done before they can actually start using agents. And I think OpenClaw was complicated. You had to get your own, you know, Mac mini. Mac minis were sold out. You had to make sure you had your own secure container. So there was a fair amount of heavy lifting, which made it good for the technically, you know, savvy people.

But for the folks that were not really spending all their time thinking about agents, this was not easy. We're now in this kind of era of personal agents, which I think have gotten the consumer market really excited. And what we'll start to see is a fair amount of kind of autonomous work being done by agents that we can delegate to the agent on our behalf. And there'll be these long-running tasks

Michael Krigsman: Mhm.

Jeetu Patel: that'll get done, long horizon tasks. And I think fundamentally this will also change the behavior of the enterprises in a big way because enterprises will now start to see a demand from the user base saying, hey, I'm using this at home, why can't I use it at work? And so we need to make sure that we have it at work. And those agents will be ones that'll be equipped with the data from work.

And so I feel like it's a very exciting time where these agents will start to do tasks and jobs for us almost fully autonomously.

Michael Krigsman: What is it about agents that will drive organizational change? Because when we talk about agentic transformation, really, that's what it is. It's like the impact of agents on the way we do work. Why? What makes this so special?

Jeetu Patel: Every workflow, every business process that you have, will actually get rethought with agents. But let's just take a step back and say, what is happening in the market right now as you make the shift to agents? And I think the statistics are really important. So right now, I would say probably best case scenario, sub-2% of humans are using agents in any kind of power user capacity.

And you'll know that even when you go to a talk and there's a few hundred people in a room and you ask them how many people use agents, there's a few hands that go up. More and more are starting to go up now with you know Muse coming out and Instinct and Grok Bot and dots and all of these. I think what you'll start to see though is that the pattern is, it's really worth noticing the pattern.

So for example, in February of this year, agent token consumption exceeded human token consumption according to a study done by OpenRouter. By August, that had gone up 14x. And so now it was like five times the human consumption of tokens by agents. Agents were now 5x that of humans on token consumption. In October, they did another study. So another quarter, and now it's 7x, the growth.

So what you're starting to see is there is this tremendous amount of consumptive pattern of agents being able to use a lot of tokens to conduct work. And 60% by the end of this year of global compute is going to be channeled towards inference, not towards training. More and more people are finding that they are delegating work to agents, and these agents are working on our behalves, and they're actually consuming a lot of infrastructure. Inferencing is the closest proxy to is there value being generated with AI. Training is not.

Training says, I'm building great models. Inference says, I'm using these models that you're building. And so inferencing is the closest proxy. So that's actually a very encouraging signal. Now, as it goes to enterprises, what we are starting to see is this tremendous need for saying, I have to move fast, and I can't just go out and have AI kind of slapped on top of existing broken processes. That's not going to work as well.

So you have to kind of rethink your processes and your business workflows within the organization to say, what do I need to do that can fundamentally transform the workflow, that can fundamentally give me a level of clock speed in the business that historically I've not been able to enjoy. And, you know, I personally feel like this notion of a legacy company is going to go away. For tech companies, for example, we will not have legacy code because you'll be able to refactor any kind of code, you know?

And so you'll constantly be able to modernize code. And so the kind of key limitations of scale go away. And the key kind of limitations of the absence of scale and startups also go away because you can actually move really fast. So I think the entire clock speed of the industry is going to be at a very different level than what we've seen ever in the past.

Team leaders should own token budgets

Michael Krigsman: So, this is from Raj Kumar, and he says he has a question about budgets, Jeetu.

Jeetu Patel: Mhm.

Michael Krigsman: And he says, "Whose budget does the token spend sit in?" Okay, we're getting down now into kind of the weeds here, but okay. When it in a central AI line, nobody in the business feels the cost or owns the result. When it sits next to headcount in the same leader's plan, the human versus agent decision gets made by the person accountable for the work. And the accountability question, he says, mostly answers itself. So, who owns the budget for these things?

Jeetu Patel: Tokens are going to be a constrained resource for a while because we are going to be in a supply-constrained world. And the reason I say that is, you know, if sub-2% of the people use agents today and we are completely fully kind of out of infrastructure available. Any kind of infrastructure that comes online gets instantly consumed. Imagine what happens when 15, 20, 30% of humans start using agents on a regular basis as power users. Boy, you're going to start to have a whole lot more consumption of tokens.

And so we will be supply constrained, in land, power, shell, compute, networking, memory for the foreseeable future. It's going to take a while for us to not be supply constrained. So then you say, so okay, so in this kind of supply constrained world, how do you dole out tokens to people and what should be the way that you think about everyone using it? My formula on this is very simple, which is this is like riding a bicycle.

You first have to get familiar with riding a bicycle before you get good at it, before you get efficient with it. So when we started using AI for coding, for example, initially for the first, you know, 4 to 6 months, I told the teams, you have unlimited use of tokens, just get fluent with AI and start using this and make sure that you don't have to be efficient. You just need to learn how to get good at it.

And you just have to get familiar with it first before you can get good at it. So just use it. Don't worry about token costs. I was a little nervous, Michael, because we didn't know how much this would actually consume in tokens. But we actually did that. Then what you started to see was the token cost started rising. People started using it a lot. And then we wanted to make sure that we actually gave them some kind of contain guardrails.

And now by the end of the month or middle of next month, we're kind of ironing out some of the kinks for you know 32,000 employees on my team. But every manager on my team is going to have a token budget, and we have a tokenomics application. That tokenomics application tells you what agents are using what level of tokens, are the agents starting to show anomalies and excessive use of tokens.

If so, can you actually pull the kill switch on that agent so that that agent doesn't actually continue to keep consuming tokens at an exorbitant rate when they're not showing the returns. And we are now starting to get into this kind of you know every manager will have a budget for what they're going to be spending for people. And they'll have a budget for what they're going to be spending for tokens. And those will be constrained resources. They're not going to be abundant resources where you have unlimited amounts.

But what you have to do is show outcomes with those constrained resources. But everyone will be assumed that they're going to have the use of tokens. And everyone's, you know, you have to assume that they will use those tokens effectively and efficiently with their team. And your ability to efficiently and effectively and securely consume tokens will be directly proportionate to not just your company's economic prosperity, but also the security posture. And I would say at the end, who owns this?

I think every company will go out and start to think about this differently, but you want to make sure that your business leaders and the team leaders are given the tooling to be able to go out and budget and regulate themselves rather than just being done at the top so that everyone's managing the budgets because that's how you create leaders in the organization.

Michael Krigsman: So, in other words, you are making or have made the transition from the experimental or pilot phase to actual, I'll use this term, production, essentially

Jeetu Patel: We're

Michael Krigsman: scaling.

Jeetu Patel: way past pilot at this point. We are way into production because the goal that we had was we are going to have half a dozen apps or products that are going to be 100% written with AI by the end of the year, no human lines of code. We will exceed that number by the end of calendar year. By the end of next year, the goal that we had was 70% of our products will be 100% written with AI.

And then I'd say within a 2-year period, 95% of the legacy code will be eliminated and refactored is the stretch goal that we've actually set for ourselves. What we had said was 100%. I think the more we think about this, is 100% going to be possible. We're going to try to push forward, but I think 95% is kind of the area that we're starting to see that we just want to eliminate as much legacy code as possible.

Token efficiency beats token volume

Michael Krigsman: Chris Petersen comes back, and he's on X, and he says, "Why is 'agents using more tokens' ever a positive metric? If the business value metrics aren't there, then it seems like there's just a ton of waste. And it seems like this is consistent.

Jeetu Patel: I'm not suggesting that agents using more tokens is a positive metric. What I'm suggesting is that is just a fact. By the way, agents are more consumptive than humans also in how they, you know, consume infrastructure. So, for example, an agent for conducting the same task as a human consumes 500x, or for 500% more network bandwidth than a human does. Why is that?

Because eventually an agent is a skills file, a markdown file, some kind of cron job, you know, and a skills markdown file, a memory markdown file, some kind of cron job and tools access.

And so when you as an agent have to keep sharing context you know on a model and if the model is not running locally, there is a fair amount of kind of upload of the model and kind of the memory and skills context that you have to make sure that you keep you know surfacing up to the agent. So I'm not suggesting that the simple use of tokens is a metric of success.

In fact, you want to make sure that you get very efficient with the use of tokens so that you don't have excessive use of tokens for an underwhelming level of return. I completely agree with that. What we are saying though is when agents start getting more consumptive in their patterns of usage, you know that there's actually more usage happening of agents and that that is the early indicator of, oh, people are starting to use agents. And that's just the kind of facts of what's happening.

And so the thing to keep in mind is there is going to be a certain amount of consumption, and that amount of consumption is causing a fair amount of, you know, kind of scarcity of infrastructure. And so what you will start to see is not just with the hyperscalers, but there'll be a multitude of different kind of creative solves for reducing the cost of tokens because the cost of tokens is going down every year, but not quite to the rate that the demand is going up for the consumption of tokens.

So if you have the cost of tokens going down 1,000x a year, but your consumption is going up 10,000x or 100,000x, you're still behind and you still are going to go out and spend more money on this. And so that's the area that we have to make sure that we actually reduce the cost of tokens sufficiently. So what are people doing?

Well, these open weights models tend to be a very you know kind of additive component because I think the more open weights models that you have that actually push to even if not frontier class, even near frontier class, not every single query and not every single action that an agent conducts needs to have a frontier class model.

And so having an intelligent routing technology directing a lot of the kind of agent tasks to an open weights model and having those open weights models sometimes run locally rather than even running in a hyperscaler or in the cloud will just allow you to balance the cost.

So I think the economics management of how many models, how you route, what specifically are the models going to be used for, where are you going to go out and run inferencing, is it going to be at your desk side computer, is it going to be local at the edge? Are you going to run it on a hyperscaler? What kind of class of GPUs are you going to run those on?

All of those things actually play the part of the calculus on how you can effectively manage token costs and still have the right level of outcome that you're looking to get from the use of the tokens. But absolutely, I completely agree with you. Lower the cost with the greatest outcome wins. You know, and each one of us has to get superbly efficient because we don't have an unlimited amount of tokens.

So when my managers get a certain amount of budget for tokens and they can get more done with a million tokens than someone else can get done with a million tokens, they end up being the better, more productive unit than someone else.

Michael Krigsman: So, you use the term "economic management," and so you are combining economic management with business outcomes, the demonstrated

Jeetu Patel: Yes.

Teams will be measured on their evals

Michael Krigsman: outcomes. All right. Let's go to another question. Oh, and I should mention, folks, that Jeetu just spoke about open-weight models. Our last episode was specifically on this topic, open-weight models, with Raffi Krikorian, who is the chief technology officer of Mozilla. So, if you're interested in open-weight models, go to CXOTalk.com and listen to our last episode because you will love it. All right. This is from Anthony Scriffignano. And he says, and he's Anthony, I know Anthony has been a guest. He's a big-time data scientist.

He says, "Can you talk a bit about regression? The ability to know or not that things still work today the way they worked yesterday while being dependent on external agentic ecosystems?"

Jeetu Patel: For example, coding as a problem is largely solved, and most code over the course of the next two, three years will largely be written with agents and not by humans. However, where I tend to have a slightly contrarian opinion to most is I actually think you're going to need more engineers, not less. And every single time there's a step function improvement in something, you know, you will have a human bottleneck that emerges. If coding becomes automated, code review becomes a bottleneck.

Code review becomes automated, you're going to have human judgment of what not to build, and which becomes so critical and so important that becomes a bottleneck. So, at the highest level, assume that coding as a problem is solved, which means over time it's going to be hard for people with the amount of code that gets generated. It's going to be hard for people to know every line of code that's written within your products.

But you're still going to need to have human provenance where every agent that writes code directed by a human needs to map back to that human to say, okay, that human is eventually accountable for that piece of code. So how do you do that? Well, one way to do it is ensure that- so assume that your code becomes a black box where you don't know what you're writing.

This is not the case today, but that's where it'll get to where, you know, you will have code that's being written to a point that it's going to be very hard for humans to inspect it. So what you will have to do is you will have to make sure that your evals are actually trying to ensure that every corner case of that code execution is happening in a way that can be tested for. And so your evals will get stronger and stronger.

And so the way that you would do it is rather than going out and looking at what the code is, what you will do is you will actually stress test that code in every condition that there is to test. And that'll actually be determined based on the quality of the evals that you generate.

And so a lot of times, organizations will be measured not just on the quality of code they write, but the quality of evals that they've actually gotten in place that can test the code that's being written so that you know what the outcomes are and not get surprised by it.

Michael Krigsman: In this world where agents are writing so much code, it becomes almost impossible to properly evaluate all of that code. So how do you do that and retain safety, security, privacy, compliance, and so on?

Jeetu Patel: Everything that you do moving forward will have to be thought about at machine scale, not at human scale, because the rate of compounding is going to be so fast that humans by themselves will not be able to do it. So for example, in cybersecurity, if your adversaries are creating attacks at machine scale and they're trying to you know find the exploit paths and vulnerabilities at machine scale, you're going to need to make sure that protecting those exploit paths and vulnerabilities getting patched is also happening at machine scale.

If you do it at human scale, it doesn't work. So, you know, evals will actually need to be- and this is why the recursive self-improvement of agents and constantly making sure that the models are keeping on getting better and better and the outcomes are keeping on learning from itself and getting better and better are going to be so important as time goes on.

But a product team within an organization will be measured just as much on the idea of the product as they are on the quality of the evals that they actually direct the agent to write for them. And then how much of the code coverage do they have and what kind of corner case testing can you do?

One of the things that we never were able to do quite right in, you know, pre-AI and pre-agents is even if you said there's X percentage code coverage with automation, you didn't quite have all the corner cases that you were able to cover. There was automation for a certain level of functional unit test, but not all the corner cases.

And now what you're going to need to have is you're going to need to keep expanding the surface area of the corner cases that you can cover, because any one corner case can become a massive security vulnerability as well. And so it becomes a security posture issue, not just a code quality issue.

Leaders need truth, not flattery

Michael Krigsman: This is from Arsalan Khan on Twitter X, who asks an interesting question here. He says, "Executives pay consultants with fancy PowerPoints to validate decisions they'd already made." Now, well, I mean, hey, he's an incisive observer of the enterprise. And so he says, "Now, AI agents are becoming the next digital transformation consultants. Will they challenge executive thinking or just automate confirmation bias at scale?"

Jeetu Patel: Yeah, are you going to just look at my watch and tell me the time, or are you going to actually have original insights? Look, I probably shouldn't say this, but we tend to have almost zero use of consultants in the product org. Within my product org because the way I think about this is if you have to ask a consultant what product to build, then I just need a different product manager, you know, because I shouldn't be asking a consultant to do that.

So I do agree that there tends to be a lot of you know show me a watch and I'll tell you the time. I think with agents, you have to make sure that the right level of harness is built where you can make sure that the agent is giving you not perfunctory feedback but actually truly substantive feedback.

And that also is largely dependent on how you go out and train and develop the agent and how you make sure that that actually gives you the kind of input that you're looking for to make you better rather than just to make you feel good. And so I do think that that's an area that I think has a lot to do with the human psychology than the agent psychology. Here's what I would say.

For hundreds of years, for as long as companies have been around, but it's gotten worse and worse over time. We as humans have started to create our own versions of the truth rather than seeking the truth when we go manage a business. You know, if you have ever encountered a time when, you know, you're looking at dashboards and everything is green, but your business is not growing at a fast enough pace.

Well, all that tells you is you might be short on ambition if everything you think is green and you're not growing the business to go out and capture share. And so I do feel like we have to be extremely, you know, kind of intentional about getting the right level of feedback from our system around us, which includes agents, which includes humans.

For example, I always surround myself with people that are going to tell me what I'm not doing well 80% of the times, and 20% of the times tell me what I'm doing well, not because I want to feel good about it, but I want to know what's working so that I can continue to keep doing that more. But you have to take away the vanity from that dimension.

And if you actually start to seek popularity in running and growing a business, I think that can become, you know, what a lot of people have said, like the desire to be liked at the expense of truth becomes a very dangerous fatal flaw of a leader. So I agree with you that we need to make sure if we are pushing our agents to give us original insights that are not you know unsatisfactory on the true alpha that's needed. You need to have high alpha insight.

Michael Krigsman: You have just described one of the key challenges that many senior leaders face. They are surrounded by people who insulate them from the harsh truths that are actually taking place.

Jeetu Patel: As the company gets large, you get very good at the math of the business. Everyone knows what your gross margin is, everyone knows what your operating income is, everyone knows all of the statistics about the business. But what you tend to lose touch with sometimes, especially as companies get large and you have more layering, is the soul of the business, which is what is the front line and the customer's view on your value, and how are you working backwards from there.

And that tends to get diluted a lot because what happens is you know, when you're a startup and you're a CEO, the CEO is the product person and is the salesperson, and they're talking to the customer, and there's no asymmetry between what the customer wants and the CEO is interpreting. They just talk directly to them, and they say, this is what the customer needs. I got this feedback, and here's what I'm going to go out and build. I'm going to take it back to them.

And there's a very fast closed loop. When it gets to a larger organization, someone talks to someone, talks to someone, that talks to someone, that talks to the customer. And before you know it, what you've gotten is it's the telephone game and then there's also perverse incentives that people have where they don't want to actually surface up the bad news. They only want to surface up the good news. And before you know it, you've lost complete touch with reality and what the customers actually want.

So it's extremely important for every leader, I don't care if you're a first-line manager or you're someone who is a senior executive at a large company, to spend some amount of time in the front lines and truly get the direct source of truth feeding you information rather than actually being fed that information via other people. Because there's a lossiness over there. The only way you can compensate for the lossiness is by going direct for some percentage of your time. You can't do that for every customer.

We have close to a million customers. I can't talk to a million customers. But I want to talk to as many of them as possible directly so I can get enough of a pulse of what they think about so I can then use that feedback in a way to direct the team in the right way. Otherwise, you're now too removed from the true problem.

Ask customers for a letter grade

Michael Krigsman: Jeetu, how do you ensure that when you talk with customers, you get the truth?

Jeetu Patel: Oftentimes, most customers might not want to offend you either, and the higher up you go in rank, the more they want to make sure that they maintain a good relationship, so they might not come out and tell you exactly how you, how are you feeling? So what I do is I usually ask some very blunt questions. One of the things that I do is I ask the customer for a letter grade.

And then I ask them, so if someone says, okay, so you got a letter grade of a C, like, okay, so what would make it an A? And then what I will do is I will typically give that customer, if there's a customer that had an escalation or they weren't happy, I immediately give that CIO or whoever we're talking to, my cell phone number and I'm like, can you just do me a favor? This will take you less than two seconds once a week.

Just send me a 1-character text message, which just gives me a letter grade. And I just want to see the progression. Is it actually getting better or is it actually getting worse? I think that that kind of complete- because otherwise what ends up happening with words sometimes is people interpret words whatever way they want to interpret words. And so even internally, we have a Customer Zero program. The Customer Zero program is every product that we build, we want to make sure that Cisco is using it at scale.

And if our own team is not willing to use it, why would a customer use it? And so oftentimes when we take that to our IT teams, our IT teams stress test it. And our CIO will typically or our head of IT will typically send me a note at the end of every week that said this week the letter grade for this as we're doing the deployment was X, and then here are the things that are missing.

And then I forward that note to my product team and I'm like, what do we need to do to get to an A and by what time are we going to get to an A? And until you get to an A, you stay relentless on that. Whatever way works for you to seek the truth, make sure that you don't penalize bad news being surfaced to you in any way, shape, or form. There should be a fully safe space for people being able to surface bad news to you.

And if they surface good news, but something is wrong, that's actually not a good thing. That means that they didn't surface the bad news to you fast enough. And that's something that you have to make sure that you have to correct right away. Because there's nothing worse than you not knowing there's a problem when there's a catastrophic problem within the organization. You have to seek those out. And you have to fix them. By the way, every company has problems.

So the fact that you have problems doesn't make you unique, doesn't make you bad. But the fact that you don't know how to fix the problems makes you less competitive. And so you just have to make sure that you find the problems and fix them, and then you find the next one and fix them, and keep doing that. And now with agents, becomes a whole lot easier to do

Human judgment cannot be fully delegated

Michael Krigsman: And this next one is from Swami Vaidyanathan, who says, "What signals warn a leader early that the agent-human mix is tipping too far in either direction? Is it escalation rates from agent to human, token spend against headcount, or something else?"

Jeetu Patel: that. I think so far, judgment, taste, intuition, deciding what not to do, those tend to all be very, very strong human characteristics and not agent characteristics. You might use an agent to help you with that, but I don't think you can actually completely delegate that to an agent. So I think there is this thing such as over-delegation to agents. One of the things that I hate is when people actually ask Claude to put together a PowerPoint deck and they don't even read it and then they give it to me.

I'm like, why? If you didn't think it was worth you reading it, why would you think that it's worth my time reading it? I want you to internalize by agents giving you data so that you can internalize so that you can apply your judgment to it. In the absence of human judgment applied, then I don't really know how valuable that is. So the way that I think about this is always look at things not getting to regress to a mean of very generic stuff.

Like one of the things right now that I'm seeing is a lot of PowerPoint decks get created by agents, and then they all start looking the same and all the messaging sounds the same and all the stories sound the same. And I'm

Michael Krigsman: Yeah,

Jeetu Patel: like,

Michael Krigsman: all the AI slop that's out there.

Jeetu Patel: It's all the AI slop and it's like, or people are just building too many products that I'm like, no one's going to use that product because that's not really a problem.

Michael Krigsman: Yes.

Jeetu Patel: I think those things, you have to have enough kind of brutal judgment applied to it from a human perspective that you don't just have unnecessary levels of genericness that gets pumped out. It's more important what you choose not to do in a world where you can do everything. Because what you choose not to do allows you to focus on doing a few things really, really well. And you know surface area is not always an indicator of success. In fact, it's the other way around.

Trusting agents takes guardrails and human owners

Michael Krigsman: Let's go to the next question. This is from Srinivas Panapakam, and he says, "It would be good to understand proactive and reactive governance of AI agents."

Jeetu Patel: Most governance over time should get proactive. Yes, of course, you need to have reactive if the agent is starting to show drift and you need to make sure that you can apply some guardrails that you want to enforce at runtime, you need to be able to intercept, one, detect that there's a drift and then be able to intercept at runtime.

But most of the times, you know, being proactive, especially in areas of outages, in areas of breaches, in areas of, you know, like agents misbehaving, those things are much, much more effective done proactively.

And there's a whole set of technologies that we're building and we're spending billions of dollars doing this to ensure that we've got the right level of ability to detect at every level of the infrastructure from the GPUs to agents to applications, what is starting to show drift that's not misaligned with human intention, and then make sure that you can intercept it right away so that you can be proactive about the damage that could be caused.

Michael Krigsman: Recently, we had the head of engineering and CIO of Snap, and as a guest on CXOTalk. And folks, you can just search for that on our site, and he said that, I think the number was 40%, may have been 50, of their engineering budget is applied to guardrails and ensuring alignment, Jeetu, as you were just describing. It's a tremendously high number.

Jeetu Patel: I think the safety and security side of the house, which is one of our big businesses, is in my opinion, the issue of our time right now, because you cannot have successful delegation to the agent if it's not trusted delegation to the agent of a body of work. All of the people that are not using agents today to the fullest capacity are because they don't trust the agent. And there's a direct correlation between the amount of tools access you give an agent and how effective that agent will be with you.

And most people don't give it enough tools access because they don't trust it. We need to get the entire agentic era to a state where people trust it and that trust is warranted because the agents aren't going rogue. And that there's enough guardrails in place that the agents are conducting what you want them to conduct on your behalf, but not doing anything that's actually harmful. And that's a very kind of you know difficult problem to solve. But there's a lot of good minds working on this right now.

Michael Krigsman: And I should mention, the context of what you just said, you know, over the last couple of days, this whole

Jeetu Patel: Yeah, the

Michael Krigsman: PHP,

Jeetu Patel: PHP stuff, yeah.

Michael Krigsman: yes, and announcements by the NSA that you need to take your sites offline, harbinger of things to come. All

Jeetu Patel: Yes.

Michael Krigsman: right. Very quickly then, a question from Hamed Zolghadri, who says, right on this point, "How do you think about risk and accountability when an AI agent takes an action that has consequences for a business? Do you see the agent as an independent entity that can be held accountable or penalized, or does responsibility lie with the engineer managing it?

Jeetu Patel: Human provenance to every agentic action is pretty important. And you have to map back to a human that actually initiated something because, you know, at some point in time, accountability needs to stop at a person. You can't just say, well, I'm not accountable. We just kind of told the agent to do something and it went out and did it. And in order to have the human feel comfortable with assuming responsibility is they need to make sure that they have enough tooling for going out and creating the right level of evals.

Stress test under, you know, kind of corner case conditions. And this is something that we as a society are still learning on how to do it. But I do think human provenance is going to be very important for people feeling trusted and safe, that there's someone that is accountable for the consequences and actions that they have actually delegated to an agent.

Michael Krigsman: And Dr. PC Jain says, "As AI agents begin to communicate and collaborate autonomously, what fundamental changes do you believe will be needed in networking and cybersecurity to make this new digital ecosystem trustworthy?"

Jeetu Patel: If you think about the way that zero trust got applied to humans, zero trust was this concept of, you know, least privileged access. You give people only the amount of permission they need so that they can get something done. Zero trust was applied to humans largely around the point of access control. You can log in, prove to me that you are who you say you are, and once you are, I'm going to give you authorization to all systems and you can do whatever you want with it. That you're authorized to you know work with.

In agents, you have to be far more granular. You don't just get access to an email system. You have to make sure that you're looking at not just access control, but action control. And so there's three things you have to keep in mind for agents. You have to give just-in-time permission, just enough permission for just enough time, and then take it away. Let me just repeat that. Just-in-time permission.

Give the permission when it's needed, just enough permission, no more, and for just enough time. And the moment the job is done, you actually take it

Preparing the next generation for agents

Michael Krigsman: Let's move on to a slightly different topic because we have a cluster of questions relating to education. And this first one is from Robert Vera, MBA, who says, "Jeetu, he leads the University of Silicon Valley. How should we be training," he says, "how should we be training the next generation of developers? What should we focus on?"

Jeetu Patel: away. I don't believe that not having computer science as a background is not needed. I actually feel like it adds a tremendous amount of value to understand the inner workings of how our machines work. That's number one.

I do feel that the job and the skills that you need for the next generation coming in that would be very exciting to learn are less around, you know, how can you go out and syntactically do something to how can you conceptually envision something and make sure that you have the ability to ask the right questions and direct the right level of work.

And so, you know, any kind of critical thinking exercise is superbly important to have because you will have tooling, where most of the time that we have spent historically has been around the syntactical understanding of what needs to happen, not the conceptual understanding of what needs to happen.

And now I think we are at the point where we can let go of the syntactical understanding almost entirely, but spend an enormous amount of time really kind of sculpting and honing that skill of conceptual understanding and framing of the problems and identifying the right problems and identifying the right questions to ask and making sure that you're actually steering with the right level of judgment. That I think is going to be a very, what I would pay to be 19 again, knowing what I know right now.

I would love to be 19 again.

Michael Krigsman: We have a walking, talking encyclopedia that we can apply, that knows everything in the world that we can apply to almost anything. And so, what you're saying makes perfect sense, that we need to develop the judgment and the conceptual and strategic skills to know how to apply that encyclopedia that knows the syntax and semantics of this system or that system. You know, it's like we have a calculator, so why should we memorize, you know, a 10-digit math table.

Jeetu Patel: But you have to make sure that if you have all of this, you know, kind of intelligence, how do you glean insight from that intelligence? That's the X factor now. And so I think that's what we have to make sure that we get doing. I don't brute memorize something, but what meaning can you extract out of that that's unique, and how can you create asymmetry for yourself and for the organization you work for is going to be a superpower skill that we all have to build.

Michael Krigsman: And on that topic, we have another question from Benyawarath Nithithanatchinnapat, who is an MIS professor at Winona State University. And I recently spoke to her class. It was very interesting. And she says, "As AI agents take on more entry-level work, how should universities rethink business education to prepare future managers and leaders.

Jeetu Patel: This notion that as AI agents take on entry-level work, you don't need entry-level employees anymore is completely bonkers. I do not subscribe to that notion at all. In fact, Cisco, we are accelerating the amount of early in career hiring that we are doing. And I feel like it's actually extremely valuable to have people with that kind of AI native mindset.

And the thing that we've found is combining the early in career with extremely experienced people is super valuable because one what you get with experience is pattern recognition, but what you also get is a liability of bias. And so you have to get really good in this era at unlearning just as much as you have to get really good at learning new things. Learning new things requires unlearning old patterns.

And what the early in career needs to do is very quickly understand patterns much faster than what they could have understood otherwise. And the combination of those two creates magic. And so one, I feel like, yes, AI will do a lot of the grunt work, but that means that even the entry level can start doing strategic work much, much faster. And what should you actually teach in school?

I think one of the most important things you can teach is the quality of questions you can ask, the kind of meaningful insights you can extract when you have all the data, and the ability to unlearn at a much, much faster pace, and not constantly say just because this worked in the past, it's going to work again in the future.

Those things, I think intrinsically people have to learn by the time they get out of you know, kind of any kind of vocational training or kind of schooling to come into the workforce. It's a superpower. If someone just is completely open-minded to have a very strong opinion that's loosely held that can be changed with new information, new data, where they can flip, I think that's a superpower. Most people think of that as like, oh, you said something five years ago, so you're going to change your mind now.

He's like, of course, when new data comes in, I change my mind every minute. You know, that's a good thing.

Product skills for agent-heavy teams

Michael Krigsman: This is from Sirj K., who says, "For people moving into product roles from other industries, which skills matter more in an agent-heavy team: domain judgment, technical depth, or something else?"

Jeetu Patel: Neither domain judgment or technical depth in my mind are the most important thing. I think the ability to have pattern recognition, the ability to be able to stitch your experiences that you've had from multiple domains together into something that brings about a unique perspective, your ability to be able to listen to and be able to decipher what the true problem is that the customer might have that they feel is going to be worth paying money for to have it solved for. Super important.

Your ability to be able to think about things where you can create an insertion point and have asymmetry. One of the worst things you can do as a product leader is say these are the things my competitors are doing, so I'm going to go out and replicate that. No one has ever won by replication. The world does not need another.

If you have competitor A, the winning formula generally is not, I'm going to just go out and replicate competitor A, just, you know, now by the way, if there's some companies that do a good job with fast follows because they have a distribution moat or something of that sort. But in general, if you don't have a distribution moat and if you're a startup, asymmetry is a much better way to do things.

And so I would say that the kind of skills I would look for are around critical thinking, around asking the right questions, around being able to extract the essence from what someone is saying in a much, much better way than what historically you might have done otherwise, and to be able to use large amounts of data and analysis and be able to find the needle in the haystack.

People are always impressive when they hear a lot of things and then they actually just distill it down to this one thing and that happens to be the thing that you need to be solving as a problem. That is an art form. The more we can actually train our minds to become good at that, the better we are going to be at having an impact on the world.

Michael Krigsman: I would also add communication. You have the distillation, and then you need to be able to share it.

Jeetu Patel: Yes. And by the way, on that one, Michael, the way I think about this is clarity of thought is extremely important because if you have clarity of thought, clarity of communication generally follows. But if you don't have clear thinking, it's impossible to become clear at communicating. And so, yes, you have to be articulate. Yes, you have to use language to your advantage.

But having clarity of thought and distilling down the core and then being able to communicate so that you can inspire people to follow you and do things that you want them to do because they want to do it, not because you told them to do it. Superbly important. And then the last one is be decisive. Make sure that you don't sit on the fence for too long. At some point in time, indecision can be paralyzing.

Michael Krigsman: Can I just say that, to me, this also links back to what you were discussing earlier, which is a ruthless search for the truth of a situation.

Jeetu Patel: I'm actually blown away at how many people aren't obsessed about finding the truth.

Michael Krigsman: It

Jeetu Patel: You

Michael Krigsman: amazes

Jeetu Patel: know.

Michael Krigsman: me too.

Jeetu Patel: Yeah.

How agents change CIO and engineer roles

Michael Krigsman: Do you have advice for CIOs, and how does the CIO role change as we go through this agentic transformation with tentacles in so many different

Jeetu Patel: Well, I think,

Michael Krigsman: areas?

Jeetu Patel: by the way, the CIO role over the years has changed quite a bit, and it had started off as a relatively tactical function that has now become not just extremely strategic, but meaningfully course-altering to a company. Let me just kind of give you this. If all of us end up using agents and these agents conduct tasks for us where they will go to a website, they will call my cable company and cancel something, they will call my retail company or go shop for me.

Up until now, what we have done over the past 30 years or so is direct-to-consumer businesses specifically have decided that they're going to be the ones that build up a lot of kind of capability, such as I'm going to have an advertising engine, I'm going to figure out cross-selling, I'm going to figure out upselling so that I can communicate directly with my customer and influence my customer to take certain actions that are going to be good for the customer, but of course eventually good for my company because if I'm a retail outlet and someone wants to buy a toothbrush, I'm going to also suggest to them a toothpaste and I'm going to suggest to them dental floss and I'm going to suggest to them an electric toothbrush.

And before you know it, you've actually increased the amount of gross merchandise volume that that retailer would be able to sell that consumer. Now you have an agent that's maybe going to shop on behalf of that consumer over time.

And as that happens, the business models might shift in a much more radical way than we think right now. Because you might not be able to have a direct line to your consumer or your customer, and you might have an intermediary, and that intermediary is aggregating the data and then providing it to the consumer. So the business model shift is non-trivial. And just like when ChatGPT came out, every company said we need to have an AI strategy. With agents, everyone's going to need to have an agent strategy.

And I think the CIOs will be the shepherds that lead the way in making sure that they can actually help their organization in figuring out not just what the technology is, but what the structural shift in the business model might end up being. And the CIO will play a very, very active role in that. So I feel like this is an exciting time to be a CIO.

I feel like this is an exciting time to participate in very strategic discussions with the business and steer that conversation because a CIO has a very unique seat where they actually get to see the technology coming in play as well. And especially for companies that are not tech-forward companies, boy, that becomes a very important role because oftentimes, if you're a nuts and bolts manufacturer, not everyone is tech savvy. But eventually every company is going to be an agentic company and the CIO might be the one that leads the way there.

So I think it's a very important role. It's one that becomes a close confidant to the CEO and will help the CEO in making the right calls and the right decisions. So, I feel like it's an exciting time for the CIO. But the CIO is also going to be responsible in working with the CISO, whether it's in their organization or someone else's, to say, what do we need to do from a safety and security and governance standpoint?

Because knowing what we have in the environment, securing and keeping safe what we have in the environment, and then governing it and auditing it are going to be baseline requirements that if we don't do well, then I think you know most organizations won't be able to get to the full potential with AI.

Michael Krigsman: And to me, this also implies that the CIO has developed those skills and judgments you were describing earlier

Jeetu Patel: Absolutely.

Michael Krigsman: because, if not, they're technicians. One last, very, very fast question because this is from your Cisco community. And this is from Yasser Ouda who says, "What AI skills do Cisco engineers need most now?"

Jeetu Patel: I would say three skills. One is let's not pay too much attention to experience and learn it all. And so the ability to learn at an accelerated pace and having that become part of the basic instinct on how we do things, super important. I think we could all continue to keep doing better at that. And we need to make sure we do that. That's number one.

Number two is every engineer will become a forward-deployed engineer because you're not going to spend time writing syntax, but you will spend a lot more time with customers. And making sure that you can help the customer understand what you're doing and how you can make that customer successful. So learning that vocation of being a forward-deployed engineer rather than just an engineer is going to be pretty important.

And then the third thing is your core ability to know that just because you're using some technology that doesn't work today, your mental model has to be, you have to assume that within the next 2 to 3 months, it's going to get working. Not within the next 9 to 12 months. So you can't shelve something that doesn't work today for 9 to 12 months and say, I'll come back to it later. You have to make sure that you assume that it's going to get fixed.

And then what you have to do is your planning cycle for going out and using that next wave of technology is going to be much longer than the actual technology getting ready and available. And that is very counterintuitive for a lot of people still, because if they try something, it doesn't work, they're like, yeah, I tried it, it doesn't work. Now I'm going to come go back to the way I was doing things. It's like, no, you have to make sure that you push the system and get there as fast.

And then the last thing I'd say is just unlearn, unlearn fast. Because your old patterns that we might have accumulated over time, especially the ones that are experienced, you have to make sure that you don't kind of hold on to them too dearly. And strong opinions loosely held, ideally that are going to be changed by data.

Michael Krigsman: Wise advice for any technical person. That was Jeetu Patel, who is Cisco's president and chief product officer. Jeetu, thank you so much for taking your time to speak with us today.

Jeetu Patel: It was such a pleasure. Thank you for having me, Michael.

Michael Krigsman: And thank you to everybody who watched, and you guys who asked questions, you are so awesome! Now, go to CXOTalk.com and subscribe to our newsletter so you can join our community. And we have amazing shows coming up, and join us! Thank you so much, everybody, and I hope you have a great day!