Operationalizing enterprise artificial intelligence requires governance trust and agentic workflows to deliver scalable service automation. Murali Swaminathan of Freshworks, chief technology officer, speaks at Refresh 2026 with Bob Laliberte of theCUBE Research, principal analyst, to discuss how organizations move AI from pilot to production.
The conversation covers governance and trust, multi-level guardrails at the large language model agent and data layers, telemetry for predictability, prebuilt AI agents and workflows, integrations with IT operations management and IT asset management and Freshworks innovations such as AI Agent Studio and MCP Gateway.
Key takeaways include the shift from experimentation to operational deployments and the necessity of layered trust traceability and data sovereignty. Swaminathan explains that anonymized training data telemetry and clear guardrails enable predictable automation and significant deflection improvements, with some customers and internal teams reporting up to 90 percent deflection. Laliberte highlights growing adoption of experience level agreements to measure employee experience alongside traditional service level agreements.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Freshworks Refresh 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Freshworks Refresh 2026
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Freshworks Refresh 2026.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Freshworks Refresh 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Freshworks Refresh 2026
Please sign in with LinkedIn to continue to Freshworks Refresh 2026. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Murali Swaminathan, Freshworks
Operationalizing enterprise artificial intelligence requires governance trust and agentic workflows to deliver scalable service automation. Murali Swaminathan of Freshworks, chief technology officer, speaks at Refresh 2026 with Bob Laliberte of theCUBE Research, principal analyst, to discuss how organizations move AI from pilot to production.
The conversation covers governance and trust, multi-level guardrails at the large language model agent and data layers, telemetry for predictability, prebuilt AI agents and workflows, integrations with IT operations management and IT asset management and Freshworks innovations such as AI Agent Studio and MCP Gateway.
Key takeaways include the shift from experimentation to operational deployments and the necessity of layered trust traceability and data sovereignty. Swaminathan explains that anonymized training data telemetry and clear guardrails enable predictable automation and significant deflection improvements, with some customers and internal teams reporting up to 90 percent deflection. Laliberte highlights growing adoption of experience level agreements to measure employee experience alongside traditional service level agreements.
In this interview from Freshworks Refresh 2026 in New York City, Murali Swaminathan, chief technology officer of Freshworks, joins theCUBE Research's Bob Laliberte to discuss how enterprises are moving from AI experimentation to trusted, production-scale operations. Swaminathan notes that enterprise AI has crossed from novelty into operational reality, with customers like Amerisure saving 4,000 hours of IT work through AI-driven automation. He outlines Freshworks' multilayered trust framework, which spans LLM-level guardrails, agent-scoped controls, data anon...Read more
exploreKeep Exploring
What changes have you observed in the enterprise AI landscape over the past year, and are organizations moving their AI projects from pilots into production?add
How does Freshworks approach responsible AI development and deployment?add
What types of tasks should be automated with AI, how should organizations measure and tune the quality of that automation, and what levels of deflection/resolution can be expected?add
>> AI is rapidly shifting from experimentation to operational deployments, and that's forcing enterprises to really rethink their governance, reliability, and service delivery. At Refresh 2026, Freshworks is highlighting how organizations can operationalize AI while maintaining trust, predictability, and enterprise scale control. Hello and welcome. We're at Refresh 2026 here in Hudson Yards in New York City. I'm Bob Laliberte, Principal Analyst for theCUBE Research, and I'm here with Murali Swaminathan, CTO of Freshworks. Murali, welcome.
Murali Swaminathan
>> Hello, Bob. Thank you. Thanks for having me.
Bob Laliberte
>> Yeah, absolutely. So we got a lot of ground to cover. I wanted to start at a high level and talk to you about from your perspective being a CTO, what have you seen change the most over the last year in enterprise AI landscapes?
Murali Swaminathan
>> There's a lot of change that has happened in the AI world. There's so much proliferation of AI native tools, third-party products. Everybody's doing AI in a different way. So it's the amount of AI that's available for you to consume, the interoperability of doing things. There's no one size fits all. So everything needs to connect together. We need to interoperate with other AI systems and that's the learning lesson here.
Bob Laliberte
>> Yeah, absolutely. And I'm wondering, are you starting to see, as I mentioned in the opening, that organizations are accelerating beyond their pilots and moving into production with AI as well?
Murali Swaminathan
>> That is right. Definitely the trust is building in. It's no longer a novelty. People are actually trying to do real things and deliver outcomes and ROI. And that's exactly what we're seeing with some of our customers are using AI. So they're actually using it to run real life production scenarios. Yeah, we have some customers who are actually using it like Amerisure, which is using AI to transform their IT. 4,000 hours of IT has been saved by using AI things. So there are such customers are doing it every single day.
Bob Laliberte
>> Awesome. No, I think that's great. I think one of the things that I wanted to touch upon is that, obviously, this AI is going to bring a lot of innovation, but they also, as you mentioned, they need some governance and trust. I believe trust was one of the big topics at Davos this year with AI. How is Freshworks approaching responsible AI development and deployment?
Murali Swaminathan
>> That's right. So without trust, there is no real action. Trust has to be in any software that you build. That's how you get the confidence that whatever the AI is doing, you're delegating your work to AI. And naturally, you want to make sure that it's doing it right way. And the trust comes in different ways. The trust in how you set up the AI, how you set up the controls, how it performs, how you can trace what's going on, and then how we can report on what it did and what it did not do. So trust is there in many layers and that's what we believe in building the product or providing the tools for you to set it up the right way the way it works for you. Yeah.
Bob Laliberte
>> Got it. So you basically enable organizations to have the guardrails to operationalize all of that governance?
Murali Swaminathan
>> That's right. So we let you set it up at different levels. It could be set up at the LLM level where you can set up guardrails for you to say what's allowed, what's not allowed. We can set up guardrails at the agent level on making sure that the agents are doing, it's an IT agent, it's only doing IT tasks and not something else. You can set it up at the way the data is trained, we anonymize the data and then we only use anonymized data. That way, it doesn't look at any personal information. There is more guardrails that are set up. We focus on data sovereignty. We make sure that the data stays in the zone and doesn't go outside. So all these are different levels of things. As far as the user is concerned, it doesn't matter to them, but behind the scenes, the people who are governing it need to know that it's working the right way. Yeah.
Bob Laliberte
>> Yeah. And that's really important, especially there's so much talk about data sovereignty these days, being able to keep that isolated and in their whatever domain they want to keep it in, so I think that's great. One of the things I also want to talk to you about is the predictability and quality. So you've talked about systems of predictability and quality. What does that mean in practice for an enterprise IT team?
Murali Swaminathan
>> AI is all about automating what is repeatable and what you have trusted and you believe in the outcomes. And so if something is predictable, repeatable, and has been human certified, then let's go back and figure out a way to automate it. And it's not about just automating it, you also want to measure what's getting automated and that's where the quality comes into play. You are going to look at the trace some of the actions that AI has taken and figure out where did it do right, where did it not do right, and then find a way to tune it to make it work better. And that's where the resolutions and deflections matter. So we have customers who start with 30% deflection and then go to 70% deflection. So some of the AI things that we have done it this time around, we are seeing 90% deflection, our internal IT teams are seeing over 90% deflection in the AI if you set it up right.
Bob Laliberte
>> Got it. Got it. And I wonder if you could touch upon the role that all the telemetry and operational data plays in helping with that.
Murali Swaminathan
>> That's right. So it's all about AI is all about tracking, reasoning, tracing. So it definitely means that you want to be able to interject as a user, user has performed a set of actions and you want to be able to go back and look at what actions were successful, what actions were not, and then look at, "Okay, where can you make the AI do it better?" And some of these actions could be automated. So if you build the trust, the human is able to see that the AI is doing working better than the human had expected, then you go ahead and automate it fully. So you start with human assisted first and then you make it autonomous.
Bob Laliberte
>> Got it. Yeah. And I think that makes a lot of sense because there's got to be a progression. I often call it the time to comfort when you're using AI. How do you want to do it before you let it go fully autonomous? You want to be there in the loop, checking it, making sure that it's doing what it's supposed to do. And a key part of that is what you're really doing is you're automating the workflows. So I'm wondering if you could touch upon what framework you use to determine which workflows are ready for agentic AI.
Murali Swaminathan
>> That's right. So it's all about we are in the service of IT and on IT teams, which would be HR, finance, legal, facilities. So we have looked at pre-determined workflows. We have like 30 plus predominant IT and HR workflows that we have picked up, and all these are vendors of choice. It could be the Workdays of the world to, the PagerDutys, to BambooHR, to the Jamf, to whatever. So the things that customers use day in, day out. So we looked at the most common newly used tools, whether it's IT, HR, other tools, and we have prebuilt AI agents and prebuilt workflows that they can just use and set it up on the get go. Basically, help them bootstrap it faster and then get it going for their needs.
Bob Laliberte
>> Got it. And they're able to then modify those as well, right? Customize them to their specific ... Because there's going to be some general workflows and then each company's going to have their own specific ones. So it just gives them really a big head start on getting going on.
Murali Swaminathan
>> That's right. That's right. Because everybody needs an example to get started, so we provide them the right example that work and then they can set it up and then they get comfortable. They can build their own AI agents too.
Bob Laliberte
>> Right. Yeah. And so then based on the feedback, that's where you've built out the stock ones that come with the tool as well.
Murali Swaminathan
>> That's right. That's right.
Bob Laliberte
>> The most popular areas.
Murali Swaminathan
>> Yes, exactly.
Bob Laliberte
>> Exactly.
Murali Swaminathan
>> And we continue to add them. So the idea is that we have our marketplace that has hundreds of integrations. So we continue to build and customers can build their own and our partners can build as well and put it on the marketplace and customers can consume.
Bob Laliberte
>> Yeah. I love that idea of the community and people being able to contribute back to it and help out other organizations to get started on a workflow that they might not already have.
Murali Swaminathan
>> That's right. That's right.
Bob Laliberte
>> Yeah. It really helps speed up. I wanted to shift a little bit and talk about some of the announcements that you had today. AI Agent Studio, MCP Gateway, ITOM, ITAM integration on the platform, experience level agreements and so forth. How do these innovations all work together to support enterprise operations?
Murali Swaminathan
>> Yeah. So we focused on a lot of breadth and depth, and AI runs on top of it. There's no AI without the data and there is no data without the data being connected. And that's what we try to do is like we brought the data together with the ITAM integration. So we brought this company called Device42. And now, it's Device42 is natively available in the cloud and then we have integrated with our ITOM flows. So we are able to take reactive incidents to make it proactive, that way we can connect the dots better. When an issue happens, we exactly know where it happened, why it happened. And then the AI leader is something that introspect into the data and it's able to help users resolve those tickets much faster.
Bob Laliberte
>> Got it. Got it. So one of the things that intrigued me was the XLAs. And so in your role, you're going out, you're talking to a lot of other technology people. Is that something they're starting to get on board with as well and saying, "Hey, we've done the whole SLA thing. It's about how all the devices and so forth are working, how the services, but we really need to move to this experience, which is ostensibly a lot more subjective from an end user and how they're experiencing it." Are you seeing organizations starting to adopt and embrace the concept of XLAs?
Murali Swaminathan
>> Actually, they are. So IT is so used to SLAs and service level agreements, but SLAs are something they use to track their work. SLA is hard for it to comprehend to the employee, but then experience level scores are whether it's not resolving the issue. Are you happy that it resolved the right way? And usually, you provide feedback in terms of CSAT surveys and post-incident surveys. XLA is nothing but collecting the rate of SLA and the CSAT things and providing a happiness score, because IT is in the service of employees. And so XLA is a great way to measure whether you are satisfying the employees or not. And so I strongly believe that XLA is the way to go, because more and more companies are trying to make it better for their employees to work and XLA is a great way to say, "Is my score going up, or coming down?"
Bob Laliberte
>> Yeah, no absolutely. And I think being able to track that is going to be hugely important and be able to show in the demo showing that you can track up or down, be able to dig in to find out why that's happened and so forth.
Murali Swaminathan
>> It's a good way for CIOs to measure whether the systems that are putting in place are really working or not.
Bob Laliberte
>> Correct. Yeah, absolutely. So we've talked a lot about AI and people consuming it. I want to talk a little bit on the technology side, how AI is being used to lower the cost to build software, shorten the time to build software, but how do you ensure that the engineering teams are staying focused on solving the right enterprise problems?
Murali Swaminathan
>> That's right. So AI definitely with the intrusion of a lot of AI coding tools, definitely it's made it easy for us to build and faster to build, that means we are able to do more things in lesser time. That means we are able to go faster on building certain things that used to take weeks or months, now takes days So the innovation velocity has definitely increased with AI, but it doesn't change how we build. We still have to go through the same guardrails of testing, validating, making sure that the product works end to end. It's deployed the right way. The customer quality. The product quality is not affected when you deliver it to the customers. So all those guardrails still exist. So the only thing is thing is they build velocity, the innovation velocity has increased. And so there's more power to the engineers, more power to the product builders to think bigger and get it done faster.
Bob Laliberte
>> Yeah. And it's great to see. And, obviously, it's a testament just six months ago in the fall, how was it Refresh. And now, it's only what, five, six months later and all the innovation you're able to produce in that short period of time, it's great to see. And it's great to see how the impact it's having on the overall industry and how much faster innovation can occur with the appropriate guardrails in place to ensure that rigor is there.
Murali Swaminathan
>> Yeah. We're actually pushing new software every couple of weeks. Of course, there are these launch moments we want to make a big deal about it, but our engineers are pushing code much faster than we announce. Yeah.
Bob Laliberte
>> Yeah, no, I think it's fantastic. This has been a great conversation. We're getting towards the end. So I just want to ask you, when you think about the next 12 to 18 months out, what really excites you about where enterprise AI is heading?
Murali Swaminathan
>> What really excites me is we are slowly moving from human assisted to autonomous. So I see that there are N number of customer workflows that can be automated and autonomized, right? That means that it can be made autonomous. And that's what I see is like, okay, you're going to a world where you can actually eliminate your L1, maybe L2 and just focus on the most complex problems. And AI is working while you're sleeping. That means that IT can focus on the bigger problems and the focus is all automation and deployment, and empowering the employees and the IT people to do more.
Bob Laliberte
>> Yeah. No, I think the great part I liked about looking at your software and the various workflows that you have is that going from human in the loop to autonomous isn't a binary function. It's not all or nothing. It can be done on a workflow at a time. And I think that's really going to be key to accelerating the adoption, so people can see it work, maybe manually trigger the workflow first and then be able to say, "Okay, after I've seen this a number of times, I'm just going to turn it over to autonomous, provided that I still get that feedback?" Right? "Hey, this is what we did and this is what we fixed."
Murali Swaminathan
>> That's how we tell our customers, like don't get started on all the time. Start with a few things, prove it to yourself, prove it to your employees and then roll it out. Yeah.
Bob Laliberte
>> Awesome. Murali, thank you so much for joining me up here at Refresh 2026, sharing how Freshworks is really thinking about trusted, scalable AI for the enterprise. So appreciate your time. And for everyone else, stay tuned for a lot more content coming from Refresh 2026.
Bob Laliberte