Red Hat Summit 2026 | Sathish Balakrishnan, Ansible

Clips
More from Red Hat Summit 2026

Sathish Balakrishnan

VP and GM

Ansible

play_circle_outline Automation is now mission-critical and foundational for AI-driven IT operations.

play_circle_outline Event-Driven Ansible: Bridging Task-Based, Event-Driven, and AI-Driven Automation with Human-in-the-Loop Safeguards

play_circle_outline Automation Orchestrator: Long-Running Jobs, Drift Detection, Integrated Workflows, Audit Trails and Observability with ServiceNow, Splunk, Dynatrace, Instana

play_circle_outline Ansible as a Trusted Execution Layer: Guardrails, RBAC, and Domain Restrictions for AI Agents

Info

Sathish Balakrishnan, Ansible

Sathish Balakrishnan

VP and GM Ansible

search

Rebecca Knight

>> Hello, everyone, and welcome back to theCUBE's live coverage of the Red Hat Summit here at the Georgia World Congress. We are on day three. I'm your host, Rebecca Knight, alongside Rob Strechay. We're talking automation today.

Rob Strechay

>> We are. We're talking automation, edge, hybrid cloud, all of this bringing together all the different places that you're going to run cloud native, and VMs, and AI, to put it mildly.

Rebecca Knight

>> Indeed. Well, with that, I would like to welcome our next guest back to the show, Sathish Balakrishnan, Vice President and General Manager, Ansible at Red Hat. Thanks so much for coming on the show, for returning to the show, I should say.

Sathish Balakrishnan

>> Well, thank you for having me. It's always great to talk to you guys.

Rebecca Knight

>> So you are running the automation business at a really, really interesting time for an enterprise IT. Sathish, talk a little bit about how ... from where you sit, the kinds of conversations you're having with customers right now here at the Red Hat Summit that you weren't having last year at this time.

Sathish Balakrishnan

>> No. I think, Rebecca, this is an exciting time to be in automation. Right? I think we've been saying this for two years now. Automation is mission-critical, and then, we've been saying last year, automation is the foundation of AI and IT operations. Now, what agentic operations has done has proven it to be true. Right? People are realizing AI is unlocking a lot of possibilities. AI ops is unlocking a lot of possibilities, but AI is probabilistic. You don't want AI agents mucking up with your production servers ever. You never put anything that's not tested into production, and this is where Ansible comes. Right? Ansible's been doing IT operations, changing servers, patching, configured management for 10 years now. All of that knowledge becomes super useful, because AI agents can leverage that, call into Ansible with the exact role-based access, the credential management, the guardrails that you need for IT ops, and make AI help IT ops become more efficient. Right? So it is a great time to be in automation.

Rob Strechay

>> How is it helping become kind of that trusted execution layer? Because people are looking at it, and they just don't even have the ... It's not that they don't have the skill sets. They just don't have enough of them in a lot of cases. So the automation, and the playbooks, and being able to go out there and really, to your point, describe it, and then execute it, and have something watch it there for them, how is this all playing out with AI?

Sathish Balakrishnan

>> Yeah. I think it all starts with how you started building automation. Right? So automation has, I'd say, multi-modes, task-based automation, event-driven automation, and then AI-driven automation. Okay? Task-based automation is a deterministic automation that you've built for a long time. Okay? That is super important, because that's proven. That's trusted. That's something you've been using for a long time. Then comes event-driven automation, because you have partners. Yesterday, we had at our product spotlight, we had Splunk, IBM Instana, and ServiceNow talk about how Ansible is the most important trusted plane to make automation changes. Right? So all of these AI observability companies are building a lot of innovative functionality, a lot of new intelligence. How can you unlock that intelligence and turn it into action? That's where event-driven Ansible comes in, and it can unlock that. Because all those insights can be translated into action. Then, we have AI-driven automation, which is, again, super important, because AI is now really, really intelligent. In three years, AI has changed the world. Right? It is so intelligent. How are you going to help take that intelligence, and instead of somebody analyzing an issue for five days, or one week, or two weeks, you can get a root cause analysis in 30 seconds. But then, it comes with new ways of doing things, but you can't really push that into production. Right? It has to be one based out of your deterministic workflows, based on your event-driven workflows, but it still needs a human in the loop to approve it. Because you are not going to allow something that just AI created into production. If you do that, we heard about what happened to a company that lost its database last weekend. You don't want that to happen to banking records. You don't want that to happen in financial services or any regulated industry where it's more mission-critical than ever before. Right? So that's where Ansible comes in, because it provides you this canvas of ways you can do all three modes of automation in the same plane so that you build on the foundation you've been working on for the last 10 years.

Rebecca Knight

>> So automation is what unlocks the opportunity of AI, but so many companies implement it reactively after things have gotten so complicated. Why is it so important to have the foundation before the AI agents come in?

Sathish Balakrishnan

>> Yeah. Why should we have a foundation before we build a house, Rebecca? Yeah.

Rebecca Knight

>> Maybe an obvious question.

Sathish Balakrishnan

>> Yes, but no. It is not so obvious. It's a great question, because this is so important. Right? Because I think people suddenly think, and this is a question I get asked by CIOs, "Oh. AI can do everything. Why do I need automation?" Right? There are two things, right? If you know how to patch a server, do you want to ask the AI, burn your very expensive tokens, burn your very expensive GPU to patch a server, which you already know how to do, and there's only one way to do it? Why would you ask AI to find 10 different ways to do the same thing and burn your GPU costs? That's the first thing. Right? The second thing is it's probabilistic. Right? It's going to iterate 10 times, and it's going to do it 10 different ways to 10 servers. Can you deal with the variability? Right? So the way I like to think about this is with AI, the cost of creating code is zero, but the complexity it creates, because now, instead of 30 million software developers, we have 300 million software developers. I pride myself to be a software developer. Don't just tell my engineering leader that. But that creates complexity. That means there's so much more output, so much more in your operations that you have to manage. But as Matt Hicks said yesterday in the keynote, complexity increased, but no resources, and IT budget didn't increase. Right? So you have to manage more with less. How do you do that? You do that only with automation. So if you don't have that foundation, and you just want AI to do it, it's going to be a disaster. You don't want to be in a disaster, especially if you're in a regulated industry, because you can't say, "You know what? I didn't do it. My AI agent did it." That's not a good answer.

Rebecca Knight

>> Not going to cut it.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> Yeah.

Rob Strechay

>> Funny enough, you can't fire the AI agent. You can rebuild it, but you can't ... So I think that's part of getting to that trusted execution layer, as well.

Sathish Balakrishnan

>> A hundred percent, hundred percent.

Rob Strechay

>> So how do you see ... Because again, you're more than just Ansible, and there's a lot of things that go along with this as you move things out to the edge and things. There's other tools and technologies. Like MCP is involved. How do you see organizations as they look to automate and kind of have that AI ops experience all the way out to the edge? How is that all playing together? Because you're involved with edge and hybrid and not just Ansible.

Sathish Balakrishnan

>> A hundred percent. Yeah. No. I think edge is a very exciting time to be in Edge, right? Because all this ... Edge is usually a very slow-moving industry, because there's a lot of hardware costs. But there's a hardware recycle that's happening today. Right? We just had the Industrial Edge Dinner yesterday, and a lot of talk about this. Right? A lot of requests for using AI in the edge now, because there's a lot of insights there. How do we get AI into the edge? Right? The good news is Ansible with the edge is a great combination, because Ansible is a multi-domain automation tool. It can do compute. It can do storage. It can do networking. I don't know if you guys saw the LinkedIn post from Cisco yesterday.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> Cisco is now reselling Ansible automation platform. Right? Great, exciting time to be. Right? So that's a validation that Ansible is the defacto automation tool for networking. Right? That's a great thing. So we do all of these domains. So edge is another domain for us which we can actually do very well. So we talked about the Edge Manager. Right? We have a new offering that basically helps us drive fleet management of edge devices, another way to do it. It also brings in AI capabilities to the edge along with everything that we are doing. That's another thing. The other thing that we are doing to make automation much easier for customers is with our 2.7 release, we are allowing customers to bring any LLM they want for playbook generation. Right? Previously, we only had Ansible Lightspeed, which we built with IBM watsonx Code Assistant like a few years back. Now, we are extending that capability to any LLM that's OpenAI API-compatible. That's a great advancement we made. The second thing is we have launched our MCP Server. So MCP Server is GA, so you can actually now call Ansible from any of your AI tool of choice. Right? That's another thing. The third thing is we are allowing customers to bring their knowledge into Ansible so that the Ansible interactive assistant that we have can actually have context about it. So you can ask questions like, "Hey. When is my maintenance window? When can I do this change?" You can ask all of those questions. That's another important thing.

Rob Strechay

>> So that's through skills and things like that.

Sathish Balakrishnan

>> That's exactly right.

Rob Strechay

>> Yeah. Okay.

Sathish Balakrishnan

>> That's exactly right, and then, the last one is the automation orchestrator. So we are allowing a new way to reimagine how automation can be done so that you can actually have long-running jobs that can actually do things like drift detection and change things automatically.

Rebecca Knight

>> So does that solve this tension between the flexibility of AI, but the precision that automation offers, too?

Sathish Balakrishnan

>> It does.

Rebecca Knight

>> It allows teams to have both.

Sathish Balakrishnan

>> Absolutely, Rebecca. What automation orchestrator does is it allows you to reimagine all the three modes of automation, task-based deterministic automation, event automation, and AI-driven automation in one place. So you can have a workflow where you have multiple agents that are listening for events, then take action, outsource it off to an AI agent. AI agent comes back with a playbook, and then you can have a human approval. As the human approval becomes more redundant for a particular task where the AI is zeroed in on that, you can actually move that to a deterministic task so you don't, anymore, use AI tokens to find the same RCA for the same issue. Right? Then, it's only used for new issues. All of this in one pane, and it integrates with your ITSM tool of choice. So ServiceNow, for example, it writes everything into ServiceNow. So you have an audit trail. It gets events from all of your observability tools like Dynatrace, Splunk, IBM Instana. So it's one pane of glass, and it can be called from either a ServiceNow pane of glass if you want, or we can call into ServiceNow, either way. It works both ways.

Rob Strechay

>> So one of the things, and you kind of hit on this a little bit, but is critical to customers I've been talking to here on the floor, is really access management. How are you looking at access management as you move forward? Because to me, part of it is, and I loved the talk track around, "Hey. We got to put it kind of ... You got to put those guardrails around it so it doesn't go and delete your entire account, your database, or your entire cloud account, or what have you." How do you look at access management playing into this?

Sathish Balakrishnan

>> Yeah. The first thing is you don't give AI agents access to your infrastructure.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> You solve it that way. Right? You only allow them to call into Ansible playbooks that are already pre-written, that are already pre-validated and tested by your team. Right? Those are the things that can do. But even there, what we want AI agents to be careful is we want AI agents ... For example, your observability agent should not be able to call into a RHEL MCP playbook and change something. Right? You got to be careful about what it calls. So you want to assign two things. One is make sure agents can't call into infrastructure directly. That's the first thing. Second thing is you want to make sure agents can only do things within their domain of knowledge. Right? So for example, you're a RHEL agent. You should be able to do only things within RHEL, and you can't go change things in networking, or Windows, or other things. Right? That's the other way you do it. The final one is if you're ever unsure, the best way to do it is put a human in the loop. Right? Have, finally, a human solve it, because the only time that you're going to be using AI in IT operations is when you find a new problem that you haven't discovered before. Because once you identify a problem, you should be able to resolve it and put it into your deterministic catalog of actions. Right? So this is where you're also saving cost, because one of the things that is not spoken as much, but if you talk to customers, is the economics of AI. As much as I said, code creation is zero. Cost of code creation is zero, but it's really not zero. Somebody's paying for those token costs, and they are expensive. Right? So that's how you do it. Right? You want to make sure you have role-based access control. You want to make sure the AI agents don't access operations, and then you only do it for the net new problems that you've never discovered before.

Rebecca Knight

>> So Red Hat gains a lot of credibility and builds a lot of trust by talking a lot about how it is solving these problems internally, too, alongside customers. When you think about the future and when automation has caught up to the agentic AI era, what is your perspective on how your team's operations, how their jobs have changed, and what those operations look like?

Sathish Balakrishnan

>> Yeah. I think Matt Hicks said it, and I think you quoted him yesterday, I think, right, as AI is not going to take your job. Somebody using AI is going to take your job. Right? I think what we are seeing is the velocity in which we are building products, the velocity in which we are responding to customer request has just dramatically increased. Right? Right from ... For example, I'll give you some. We are talking about an inventory service, "Hey. How do we build this?" Then, three hours later, I have a working prototype of how it will look like. Right? So that is a great way for our product teams to explain to our engineering, "Here is what we are looking for." Right? A prototype can be built in three hours. A product cannot be built in three hours, because our engineering team has to worry about a lot of things about enterprise stability, the logging, the audit, the permissions that you need, the scalability, and all of those things. So we have to architect it in the right way. But what is happening is the delta between the expectation that product managers have with what engineering is delivering is actually narrowing and shrinking. Right? That's one thing. The other thing that's happening from a product marketing standpoint is our product marketers now have to catch up to the speed at which we are delivering software. But the good news is they have AI. AI can help them, "Hey. Explain this to me like you would explain it to a high schooler, and then, okay, then how would I sell this to an enterprise customer?" They can actually build these marketing collateral much more easier and actually sharpen to the focus of the persona that we are actually targeting for marketing. Right? That has improved, as well. From a technical marketing standpoint, what we do is we do a lot of demos, we do a lot of workshops. So what is happening is these prototypes, they can actually build, and they can actually start simulating things like, "Hey. Simulate that there is a network switch that you're configuring, and how do you do that?" So that gives our customers a better hands-on experience without really touching a Cisco switch or touching a GPU which may not be available. So that's how AI is accelerating the way that we are doing things in the product area, and there's a lot of other things that, as Matt described yesterday, we are using AI for our operations. So for example, one of the teams that work for me is Cloud Marketplaces. So how do I get a cloud marketplace private offer out? We can do that in like ... It used to take us like two days of working through multiple Google forms, and we do that in three steps today, all AI-driven, where we know the context of what they're asking. Somebody files a Salesforce ticket, and it automatically becomes a private offer in some ways. Right? So AI is really transforming the way that we are operating, and that's the knowledge that I think Matt and team want to bring to our customers.

Rebecca Knight

>> Excellent. Well, thank you so much, Sathish. Always a pleasure having you on.

Sathish Balakrishnan

>> Thank you, guys.

Rob Strechay

>> Thank you.

Sathish Balakrishnan

>> It's been awesome talking to you, and have a great rest of summit.

Rebecca Knight

>> Thank you.

Rob Strechay

>> Thank you.

Rebecca Knight

>> You, too. I'm Rebecca Knight for Rob Strechay. Stay tuned for more of theCUBE's live coverage of the Red Hat Summit. You're watching theCUBE, the leader in enterprise tech news and analysis.

Sathish Balakrishnan, Ansible

search

Rebecca Knight

>> Hello, everyone, and welcome back to theCUBE's live coverage of the Red Hat Summit here at the Georgia World Congress. We are on day three. I'm your host, Rebecca Knight, alongside Rob Strechay. We're talking automation today.

Rob Strechay

>> We are. We're talking automation, edge, hybrid cloud, all of this bringing together all the different places that you're going to run cloud native, and VMs, and AI, to put it mildly.

Rebecca Knight

>> Indeed. Well, with that, I would like to welcome our next guest back to the show, Sathish Balakrishnan, Vice President and General Manager, Ansible at Red Hat. Thanks so much for coming on the show, for returning to the show, I should say.

Sathish Balakrishnan

>> Well, thank you for having me. It's always great to talk to you guys.

Rebecca Knight

>> So you are running the automation business at a really, really interesting time for an enterprise IT. Sathish, talk a little bit about how ... from where you sit, the kinds of conversations you're having with customers right now here at the Red Hat Summit that you weren't having last year at this time.

Sathish Balakrishnan

>> No. I think, Rebecca, this is an exciting time to be in automation. Right? I think we've been saying this for two years now. Automation is mission-critical, and then, we've been saying last year, automation is the foundation of AI and IT operations. Now, what agentic operations has done has proven it to be true. Right? People are realizing AI is unlocking a lot of possibilities. AI ops is unlocking a lot of possibilities, but AI is probabilistic. You don't want AI agents mucking up with your production servers ever. You never put anything that's not tested into production, and this is where Ansible comes. Right? Ansible's been doing IT operations, changing servers, patching, configured management for 10 years now. All of that knowledge becomes super useful, because AI agents can leverage that, call into Ansible with the exact role-based access, the credential management, the guardrails that you need for IT ops, and make AI help IT ops become more efficient. Right? So it is a great time to be in automation.

Rob Strechay

>> How is it helping become kind of that trusted execution layer? Because people are looking at it, and they just don't even have the ... It's not that they don't have the skill sets. They just don't have enough of them in a lot of cases. So the automation, and the playbooks, and being able to go out there and really, to your point, describe it, and then execute it, and have something watch it there for them, how is this all playing out with AI?

Sathish Balakrishnan

>> Yeah. I think it all starts with how you started building automation. Right? So automation has, I'd say, multi-modes, task-based automation, event-driven automation, and then AI-driven automation. Okay? Task-based automation is a deterministic automation that you've built for a long time. Okay? That is super important, because that's proven. That's trusted. That's something you've been using for a long time. Then comes event-driven automation, because you have partners. Yesterday, we had at our product spotlight, we had Splunk, IBM Instana, and ServiceNow talk about how Ansible is the most important trusted plane to make automation changes. Right? So all of these AI observability companies are building a lot of innovative functionality, a lot of new intelligence. How can you unlock that intelligence and turn it into action? That's where event-driven Ansible comes in, and it can unlock that. Because all those insights can be translated into action. Then, we have AI-driven automation, which is, again, super important, because AI is now really, really intelligent. In three years, AI has changed the world. Right? It is so intelligent. How are you going to help take that intelligence, and instead of somebody analyzing an issue for five days, or one week, or two weeks, you can get a root cause analysis in 30 seconds. But then, it comes with new ways of doing things, but you can't really push that into production. Right? It has to be one based out of your deterministic workflows, based on your event-driven workflows, but it still needs a human in the loop to approve it. Because you are not going to allow something that just AI created into production. If you do that, we heard about what happened to a company that lost its database last weekend. You don't want that to happen to banking records. You don't want that to happen in financial services or any regulated industry where it's more mission-critical than ever before. Right? So that's where Ansible comes in, because it provides you this canvas of ways you can do all three modes of automation in the same plane so that you build on the foundation you've been working on for the last 10 years.

Rebecca Knight

>> So automation is what unlocks the opportunity of AI, but so many companies implement it reactively after things have gotten so complicated. Why is it so important to have the foundation before the AI agents come in?

Sathish Balakrishnan

>> Yeah. Why should we have a foundation before we build a house, Rebecca? Yeah.

Rebecca Knight

>> Maybe an obvious question.

Sathish Balakrishnan

>> Yes, but no. It is not so obvious. It's a great question, because this is so important. Right? Because I think people suddenly think, and this is a question I get asked by CIOs, "Oh. AI can do everything. Why do I need automation?" Right? There are two things, right? If you know how to patch a server, do you want to ask the AI, burn your very expensive tokens, burn your very expensive GPU to patch a server, which you already know how to do, and there's only one way to do it? Why would you ask AI to find 10 different ways to do the same thing and burn your GPU costs? That's the first thing. Right? The second thing is it's probabilistic. Right? It's going to iterate 10 times, and it's going to do it 10 different ways to 10 servers. Can you deal with the variability? Right? So the way I like to think about this is with AI, the cost of creating code is zero, but the complexity it creates, because now, instead of 30 million software developers, we have 300 million software developers. I pride myself to be a software developer. Don't just tell my engineering leader that. But that creates complexity. That means there's so much more output, so much more in your operations that you have to manage. But as Matt Hicks said yesterday in the keynote, complexity increased, but no resources, and IT budget didn't increase. Right? So you have to manage more with less. How do you do that? You do that only with automation. So if you don't have that foundation, and you just want AI to do it, it's going to be a disaster. You don't want to be in a disaster, especially if you're in a regulated industry, because you can't say, "You know what? I didn't do it. My AI agent did it." That's not a good answer.

Rebecca Knight

>> Not going to cut it.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> Yeah.

Rob Strechay

>> Funny enough, you can't fire the AI agent. You can rebuild it, but you can't ... So I think that's part of getting to that trusted execution layer, as well.

Sathish Balakrishnan

>> A hundred percent, hundred percent.

Rob Strechay

>> So how do you see ... Because again, you're more than just Ansible, and there's a lot of things that go along with this as you move things out to the edge and things. There's other tools and technologies. Like MCP is involved. How do you see organizations as they look to automate and kind of have that AI ops experience all the way out to the edge? How is that all playing together? Because you're involved with edge and hybrid and not just Ansible.

Sathish Balakrishnan

>> A hundred percent. Yeah. No. I think edge is a very exciting time to be in Edge, right? Because all this ... Edge is usually a very slow-moving industry, because there's a lot of hardware costs. But there's a hardware recycle that's happening today. Right? We just had the Industrial Edge Dinner yesterday, and a lot of talk about this. Right? A lot of requests for using AI in the edge now, because there's a lot of insights there. How do we get AI into the edge? Right? The good news is Ansible with the edge is a great combination, because Ansible is a multi-domain automation tool. It can do compute. It can do storage. It can do networking. I don't know if you guys saw the LinkedIn post from Cisco yesterday.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> Cisco is now reselling Ansible automation platform. Right? Great, exciting time to be. Right? So that's a validation that Ansible is the defacto automation tool for networking. Right? That's a great thing. So we do all of these domains. So edge is another domain for us which we can actually do very well. So we talked about the Edge Manager. Right? We have a new offering that basically helps us drive fleet management of edge devices, another way to do it. It also brings in AI capabilities to the edge along with everything that we are doing. That's another thing. The other thing that we are doing to make automation much easier for customers is with our 2.7 release, we are allowing customers to bring any LLM they want for playbook generation. Right? Previously, we only had Ansible Lightspeed, which we built with IBM watsonx Code Assistant like a few years back. Now, we are extending that capability to any LLM that's OpenAI API-compatible. That's a great advancement we made. The second thing is we have launched our MCP Server. So MCP Server is GA, so you can actually now call Ansible from any of your AI tool of choice. Right? That's another thing. The third thing is we are allowing customers to bring their knowledge into Ansible so that the Ansible interactive assistant that we have can actually have context about it. So you can ask questions like, "Hey. When is my maintenance window? When can I do this change?" You can ask all of those questions. That's another important thing.

Rob Strechay

>> So that's through skills and things like that.

Sathish Balakrishnan

>> That's exactly right.

Rob Strechay

>> Yeah. Okay.

Sathish Balakrishnan

>> That's exactly right, and then, the last one is the automation orchestrator. So we are allowing a new way to reimagine how automation can be done so that you can actually have long-running jobs that can actually do things like drift detection and change things automatically.

Rebecca Knight

>> So does that solve this tension between the flexibility of AI, but the precision that automation offers, too?

Sathish Balakrishnan

>> It does.

Rebecca Knight

>> It allows teams to have both.

Sathish Balakrishnan

>> Absolutely, Rebecca. What automation orchestrator does is it allows you to reimagine all the three modes of automation, task-based deterministic automation, event automation, and AI-driven automation in one place. So you can have a workflow where you have multiple agents that are listening for events, then take action, outsource it off to an AI agent. AI agent comes back with a playbook, and then you can have a human approval. As the human approval becomes more redundant for a particular task where the AI is zeroed in on that, you can actually move that to a deterministic task so you don't, anymore, use AI tokens to find the same RCA for the same issue. Right? Then, it's only used for new issues. All of this in one pane, and it integrates with your ITSM tool of choice. So ServiceNow, for example, it writes everything into ServiceNow. So you have an audit trail. It gets events from all of your observability tools like Dynatrace, Splunk, IBM Instana. So it's one pane of glass, and it can be called from either a ServiceNow pane of glass if you want, or we can call into ServiceNow, either way. It works both ways.

Rob Strechay

>> So one of the things, and you kind of hit on this a little bit, but is critical to customers I've been talking to here on the floor, is really access management. How are you looking at access management as you move forward? Because to me, part of it is, and I loved the talk track around, "Hey. We got to put it kind of ... You got to put those guardrails around it so it doesn't go and delete your entire account, your database, or your entire cloud account, or what have you." How do you look at access management playing into this?

Sathish Balakrishnan

>> Yeah. The first thing is you don't give AI agents access to your infrastructure.

Rob Strechay

>> Yeah.

Sathish Balakrishnan

>> You solve it that way. Right? You only allow them to call into Ansible playbooks that are already pre-written, that are already pre-validated and tested by your team. Right? Those are the things that can do. But even there, what we want AI agents to be careful is we want AI agents ... For example, your observability agent should not be able to call into a RHEL MCP playbook and change something. Right? You got to be careful about what it calls. So you want to assign two things. One is make sure agents can't call into infrastructure directly. That's the first thing. Second thing is you want to make sure agents can only do things within their domain of knowledge. Right? So for example, you're a RHEL agent. You should be able to do only things within RHEL, and you can't go change things in networking, or Windows, or other things. Right? That's the other way you do it. The final one is if you're ever unsure, the best way to do it is put a human in the loop. Right? Have, finally, a human solve it, because the only time that you're going to be using AI in IT operations is when you find a new problem that you haven't discovered before. Because once you identify a problem, you should be able to resolve it and put it into your deterministic catalog of actions. Right? So this is where you're also saving cost, because one of the things that is not spoken as much, but if you talk to customers, is the economics of AI. As much as I said, code creation is zero. Cost of code creation is zero, but it's really not zero. Somebody's paying for those token costs, and they are expensive. Right? So that's how you do it. Right? You want to make sure you have role-based access control. You want to make sure the AI agents don't access operations, and then you only do it for the net new problems that you've never discovered before.

Rebecca Knight

>> So Red Hat gains a lot of credibility and builds a lot of trust by talking a lot about how it is solving these problems internally, too, alongside customers. When you think about the future and when automation has caught up to the agentic AI era, what is your perspective on how your team's operations, how their jobs have changed, and what those operations look like?

Sathish Balakrishnan

>> Yeah. I think Matt Hicks said it, and I think you quoted him yesterday, I think, right, as AI is not going to take your job. Somebody using AI is going to take your job. Right? I think what we are seeing is the velocity in which we are building products, the velocity in which we are responding to customer request has just dramatically increased. Right? Right from ... For example, I'll give you some. We are talking about an inventory service, "Hey. How do we build this?" Then, three hours later, I have a working prototype of how it will look like. Right? So that is a great way for our product teams to explain to our engineering, "Here is what we are looking for." Right? A prototype can be built in three hours. A product cannot be built in three hours, because our engineering team has to worry about a lot of things about enterprise stability, the logging, the audit, the permissions that you need, the scalability, and all of those things. So we have to architect it in the right way. But what is happening is the delta between the expectation that product managers have with what engineering is delivering is actually narrowing and shrinking. Right? That's one thing. The other thing that's happening from a product marketing standpoint is our product marketers now have to catch up to the speed at which we are delivering software. But the good news is they have AI. AI can help them, "Hey. Explain this to me like you would explain it to a high schooler, and then, okay, then how would I sell this to an enterprise customer?" They can actually build these marketing collateral much more easier and actually sharpen to the focus of the persona that we are actually targeting for marketing. Right? That has improved, as well. From a technical marketing standpoint, what we do is we do a lot of demos, we do a lot of workshops. So what is happening is these prototypes, they can actually build, and they can actually start simulating things like, "Hey. Simulate that there is a network switch that you're configuring, and how do you do that?" So that gives our customers a better hands-on experience without really touching a Cisco switch or touching a GPU which may not be available. So that's how AI is accelerating the way that we are doing things in the product area, and there's a lot of other things that, as Matt described yesterday, we are using AI for our operations. So for example, one of the teams that work for me is Cloud Marketplaces. So how do I get a cloud marketplace private offer out? We can do that in like ... It used to take us like two days of working through multiple Google forms, and we do that in three steps today, all AI-driven, where we know the context of what they're asking. Somebody files a Salesforce ticket, and it automatically becomes a private offer in some ways. Right? So AI is really transforming the way that we are operating, and that's the knowledge that I think Matt and team want to bring to our customers.

Rebecca Knight

>> Excellent. Well, thank you so much, Sathish. Always a pleasure having you on.

Sathish Balakrishnan

>> Thank you, guys.

Rob Strechay

>> Thank you.

Sathish Balakrishnan

>> It's been awesome talking to you, and have a great rest of summit.

Rebecca Knight

>> Thank you.

Rob Strechay

>> Thank you.

Rebecca Knight

>> You, too. I'm Rebecca Knight for Rob Strechay. Stay tuned for more of theCUBE's live coverage of the Red Hat Summit. You're watching theCUBE, the leader in enterprise tech news and analysis.