Designing AI-Native Service Operations: From Automation to Resolution

Most enterprise service environments were built around queues, escalation paths, and fragmented workflows long before AI entered the picture. Layering AI onto that architecture delivers incremental gains at best. It rarely changes how service actually operates, and the companies seeing meaningful gains in customer and employee experience are rethinking the model itself, treating resolution as the core outcome and using intelligence to coordinate work across systems, teams, and channels.

‍

At Zendesk Relate 2026 in Denver, Keith Kirkpatrick and Melody Brue are joined by Vishnu Parimi, VP of Employee Service Management at Zendesk to explore what that shift looks like in practice. They look at how enterprises are moving beyond disconnected automations toward service architectures built around unified resolution, why specialized agents are proving more effective than broad general-purpose AI in complex service environments, and how product teams are balancing scalability, governance, and operational control as autonomy increases across workflows.

Vishnu also breaks down the operational realities that many organizations fail to plan for when AI moves beyond the pilot stage. The challenge is rarely the model itself. It’s the coordination overhead, process redesign, governance structure, and organizational alignment required to make AI-driven service systems work reliably at scale.

‍

Key Takeaways:

‍

🔹Legacy service stacks break under agentic AI. Rule-based routing, isolated automation, and siloed workflows create structural friction that prevents AI from delivering consistent resolution at scale. AI-native design requires rethinking the operational model before layering in the technology.

‍

🔹Unified resolution layers outperform disconnected automation. Point-solution approaches can’t replicate the higher completion rates, fewer escalations, and better customer outcomes enterprises see when systems work together.

🔹Specialized agents beat general-purpose AI in production environments. Service workflows requiring policy application, contextual memory, and domain-specific judgment deliver better outcomes with purpose-built agents than with broad-purpose models deployed across every interaction type.

🔹Governance becomes critical as autonomy increases. The risk of inconsistent customer interactions, policy violations, and compounding errors grows significantly as agentic systems expand. Product teams must build observability, control, and correction mechanisms into the architecture from the start.

‍

🔹Most organizations underestimate the operational lift required to scale AI. Technical deployment is the smallest part of the challenge.Workflow redesign, quality monitoring, and organizational alignment are where deployments stall and where investment must be front-loaded.

‍

🔹 AI-native service is not a feature upgrade on top of existing operations. It is a structural redesign that demands as much organizational investment as it does technological investment.

Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel so you never miss an episode.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.

VISHNU PARIMI:
The most complex ones are probably you haven't seen before. And that's where you make a customer really, really unhappy. And you really need to know what are those situations, and within your AI solutions, be able to define those escalation paths.

MELODY BRUE: 

Hello and welcome to Six Five On The Road. I'm Melody Brew and I am joined by Keith Kirkpatrick and Vishnu Parimi and we are at Zendesk Relate 2026 here in Denver. Vishnu, thank you so much for joining us.

VISHNU PARIMI: 

Thank you for having me as a guest.

MELODY BRUE: 

Yeah, big day today, big exciting keynote, lots of announcements. Before we get into a bunch of questions, I want to just ask you, one of the things I wasn't expecting was this focus around employee service. Let's talk a little bit about that and why that's important right now for Zendesk.

VISHNU PARIMI: 

Yeah, a lot of people don't know about it, but when we talk about service, it's always about customer service. But equally important and equally big market is employee service. Happy employees lead to happy business outcomes, lead to happy customers. So it's kind of a loop. So it makes very much sense for Zendesk to be concentrating on this because it's still service at the end of the day. Now, why now and why it wasn't like a decade earlier? I think to an extent there's the macro phenomena with COVID employees not coming to office anymore and then they're working in a distributed fashion. So what happens as a result of service? An employee, instead of coming and asking about his laptop challenge to an in-office help desk, you are now asking about these questions more remotely. So how do you actually provide this service? You need a solution like customer support to be able to also deliver equally empathetic employee support. And it is a big market. I think there are players like ServiceNow and others that have done really well in this market. And we are also seeing players like Salesforce recognize the total market value of this opportunity. And that's one of the reasons Zendesk is also heavily investing in it. Our customers are asking, without even building employee support, we already have a sizable business there. So it made all the more sense for us to start investing in it and grow our business in this way.

MELODY BRUE: 

It also makes a lot of sense to me that people want consumer-grade tools at work. So for an employee, as they're used to kind of interacting with the Zendesk brand when they're shopping or traveling or doing something else, that makes it a really easy extension for them to do that at work as well, I would think, right?

VISHNU PARIMI: 

I think that's the expectation we are having, like when employees interact with other brands, I think the question to them internally is, you know, when I'm interacting with my own brand, with my own company, should I be getting the same level of support? I think that's the expectation, and that's the expectation more and more as we actually build these AI-native solutions that are probably giving instant support to employees. And as a company, you have to manage those expectations. And it can be not just your existing employees, right? You're getting employees from different organizations that are probably providing that level of support. And when those employees are coming and joining you, they would expect a similar kind of support in terms of enabling them to do their best work. Ultimately, probably 60, 70% of their life is probably their work life. So you want to make sure that they're taken care of while they're doing work.

MELODY BRUE: 

Yeah. That employer brand it's important.

KEITH KIRKPATRICK: 

Yeah. So Vishnu, you know, as organizations really kind of accelerate adoption across different customer experience environments, what is some of the organizational challenges that often occur with legacy environments?

VISHNU PARIMI: 

Yeah. With legacy environments, you often see that, you know, it takes a long time to implement. They carry a lot of load. And as a result, you will see that there is ballooning costs ultimately over time because initially it might be very inexpensive. The secondary thing is then you have to bring in someone to build on top of it because it's legacy. It's no more built for AI. And then those implementation and maintenance costs become unbearable over time. And that's some of the challenges we always see with our existing customers today, where they're coming from other companies like ServiceNow, and then they come to Zendesk, and they say, oh, great, your solution is so easy to use. And it's also AI native and AI ready. And we have really gone from paying a lot for implementation, but still not getting the service on time, to a very cost-efficient solution, but also very AI-native and ready to actually provide service within seconds.

KEITH KIRKPATRICK: 

Right.

MELODY BRUE: 

So one of the things that you mentioned was the speed to implementation, but this requires a long-term, like learning over time, systems that learn with what the customer is trying to achieve. What does that take from a product perspective, and how do you build on that sort of adaptive operational model?

VISHNU PARIMI: 

As we can see, Zendesk, as a service system, whether it is customer service or an employee service, they actually collect a lot of data about customer pain or employee pain. And this wealth of data is only available with support systems like us because it is an enterprise-focused data that we store, and it's not available for any outside LLMs to be able to glean at it. So what do we do with this row of data that we have within our system? So previously, we did not have the technology to be able to read all this unstructured pain to be able to figure out, hey, what are the three big drivers of this pain that we can quickly address that will help organization provide better service? Now with generative AI, being able to train and really learn on this data and be able to read, and it's really good at giving those insights. Now, we are able to identify those and provide it to customers out-of-the-box, and that's where the ease of value with Zendesk comes in. Once they see this, it's like, hey, we already have all this, why don't we take action on it? Action can be taken two ways. Action can be taken through tools that we are providing to automate their processes, either through AI, or you can look at some of these things as like these are unusual cases, so what do I do about it? then you can actually bring in the human touch and your human workforce to help and provide that empathy. That's where you provide a mix of automated and empathized support to actually deliver the best experience for customers and employees.

KEITH KIRKPATRICK: 

One of the things that we've been hearing a lot about this week at Relate is this idea of specialization. I was wondering if you could talk a little bit about why using a more specialized approach makes more sense in this sort of generalized AI in terms of 

driving real benefit.

VISHNU PARIMI: 

Yeah that's a good question because when we started so most of our AI solutions when we started were very similar whether it is a customer support or an employee support use case right but then we started realizing because we got started getting a lot of customer feedback around it saying that Hey, let's take an example of an employee service use case. You are an employee and you're trying to get some data about a policy of a particular company. Some of these data are behind systems that are authenticated. So if employee asks saying that, hey, I need to know what my termination policy is in this company, they shouldn't be able to view it. It's only locked to HR people to be able to internalize it and make it a process within a company. So those are not common things when you see in terms of customer support. So when you're building AI for employee support, you need to actually make sure that you are governing all those access controls that you have with internal content before surfacing it to your employee. So that's why these verticalized solutions will be better to be able to solve those use cases that are very specific to that vertical. But having said that, not everything is vertical, right? Like the previous example that we talked about. you know, we have pain. You have to learn what the pain is and then highlight, you know, what are the three drivers of the pain. That's a common use case. And that's where Zendesk platform is able to scale for both service use cases. But on top of that, we are implementing or operationalizing these specialized agents that will be able to address the nuances of each of these verticals.

MELODY BRUE: 

So that really becomes more of a multi-use agent across the verticals. And in that you really do have to balance those permissions, the guardrails, the governance. So explain a little bit how organizations should be thinking about maintaining those guardrails and that trust. while also getting to that place of automation where you're talking about where you have thousands if not hundreds of thousands of tickets and this requires some level of automation that there's a trust factor still that the industry is kind of grappling with.

VISHNU PARIMI: 

Yes, yes. I think this is the number one question in everybody's mind is right now. I think this is where I think the, I would say the magic of AI meets the joy of operations.

MELODY BRUE: 

Those are both positive things. You did that well.

VISHNU PARIMI: 

So it's just that if you balance both of them well, then you can actually see this autonomous world that AI is promising. So let me give you some specific examples, right? When a company is implementing AI, it has to also think about, how do I govern the AI that is acting on my data? So a couple of things we have already talked about, like security, permissions, which content does AI access to? And all these governance should be in place before even deploying your AI solutions. That's number one. Number two I see is, you know, AI is able to do more, and that's why you're deploying AI, right? Previously, if your human workforce was able to look at, let's say, 100 interactions, as an example, AI could probably do 1,000 of interactions. So that means somebody has to actually go review those interactions and see if it's the same quality and it's being offered with the same consistency. how do you actually do it with scale when the number of interactions you are having is 10x today? So that's another thing that organizations need to really think about when they are deploying it at scale. The final thing I would say is, there's always these situations, like I said, are unusual situations. What I say is, generally, if you see in support, the pain is always on the edges. By that, I mean that, The most complex ones are probably you haven't seen before. That's where you make a customer really, really unhappy. You really need to know what are those situations and within your AI solutions, be able to define those escalation paths. Hey, this is probably not a common situation, so maybe a human could come in and help here and probably be more empathetic than what my policy says to be able to address their pain.

KEITH KIRKPATRICK: 

So kind of looking ahead, what is, you know, what are some of the things that are, I guess, underestimated by organizations as they start to affect this change? And what should they really focus on, you know, to make sure that the transition is as smooth as possible?

VISHNU PARIMI: 

Yeah, I think there are a couple of things that come to my mind, right? When they look at these transitions, first they have to believe that AI is not there for incrementally assisting to actually for somebody to do their job. It is probably also be able to completely automate some chunks of it. When you actually believe in that philosophy, then you will see the maximum benefit of deploying these solutions. Now, as a result of that, you will have other things to look at to actually go to this promise world. Things like, hey, how should your organization look like right now? Previously without AI, you were handling as humans a lot of those interactions. So then there were only a few admins and then a bunch of work force that is just directly communicating. But now, if you're deploying AI, then somebody has to build those and somebody has to manage those. And probably you need a different mix of talent to go to this world. Now, what would that right mix be? Maybe it could be a mix of more admins to be able to deploy this technology and lesser folks that are directly interacting with the customer or the employee. So they have to think about the balance of their organization in this future world. So when you actually look at both of these, these are the organizations you see that are successfully navigating this path of going into autonomous operations.

MELODY BRUE: 

Before we wrap up, we talked a lot about people have to believe, they have to trust. Give us some reasons to believe. What are the outcomes that the customer sees, whether that's customer satisfaction, that's revenue driver, where you're seeing the real wins that give people a reason to believe?

VISHNU PARIMI: 

Yes, we should always think why we are deploying AI solutions. If 

ultimately we have to solve customer pain. I'll give you an example, like how many of us like have seen airline disruptions, many of them during our lifetime, right? Like most recently, I was planning to travel to India and I booked an Emirates flight and I don't know whether it'll go or it'll fly because of all the Middle East conflict. Now, in such cases, what happens is you'll get a deluge of customer requests on what's happening for a particular company. And no matter how well you plan your workforce for the entire year, when you hit those spots, you're not going to handle the traffic. Now, I think in this newer world, you would be able to do it. And the reason you would be able to do it is you can deploy a lot of this automation so that customers, when things go really bad for a particular company, and when they're calling for support, you're not making them wait 24 hours, 48 hours to get support. you're instantly handling their request during the times of their utmost pain and utmost need. That's where I see this going. For a particular company, the most negative PR around service comes during these moments. If AI can really help eliminate all this, that's truly beneficial both from a customer perspective and from a company's perspective.

KEITH KIRKPATRICK: 

Well, that gives us a lot to think about. So, thank you very much Vishnu. Thank you very much Mel. And thank you for tuning in to Six Five On The Road at Zendesk Relate 2026. Don't forget to hit subscribe, check out the socials, and check out all of our coverage at SixFiveMedia.com. Thanks, and we'll see you very soon.