The Unified Ops Engine

Karan Munalingal • 00:09

Good afternoon, and welcome everyone to our webinar that we’re exclusively hosting for Verizon and everyone participating from Verizon. Today’s topic: we’re going to be covering how the Atential platform and its innovation can accelerate the Verizon’s 2026 objectives in support of business growth and operational efficiency. So in today’s webinar, we’re gonna be covering a few set of things. Let’s talk through the agenda. We’re gonna do some quick introductions for the team that’s gonna be leading the webinar today. Uh, we’re gonna do some intentional overview and the evolution of the product portfolio. We’re gonna be discussing a little bit about our partnership with Verizon for the past eight plus years.

Karan Munalingal • 00:54

Uh, and then we also have some goodies. We have two demonstrations today being that’s gonna be done by Joksan and Dalton in support of concepts around stateful orchestration and how that will be beneficial to Verizon in various different domains, as well as agentic network operation. And AI is the new hype. So we’re gonna be uh showcasing some capabilities on how platform capabilities could be paired up with reasoning as well as AI up top in order to consume infrastructure services safely. So let’s talk a little bit about Itential. Who are we? For anyone who may not know, uh, we are an orchestration platform with software company uh focused on providing orchestration platform for the AI era, right?

Karan Munalingal • 01:37

From our standpoint, we have been doing this for a long time with Verizon and other service provider customers, uh supporting multi-domain and vendor agnostic automations and orchestration, as well as being able to integrate across all your uh investments within the service provider space. But at the same time, we have done this at speed and scale. Ittential is known for its scale with a lot of our customers, similar to Verizon, but specifically within Verizon. And finally, the way that we have built our platform uh makes it AI ready, right? When you think about some of the capabilities that we’re gonna talk through today that Verizon already leverages, it basically gives you a leg up with respect to how do you go to market faster with the AI capabilities up top. So, how does it look drive business value for service providers, right? In general, but also for uh Verizon and how we can continue doing that into 2026.

Karan Munalingal • 02:36

On the left, you see revenue generation, right? Number one, I think it’s it’s very critical as a platform when we integrate, when we automate, is in order to help drive and reduce the activation time. This is how your customers get their services that they’re procuring faster. So from a time to value standpoint, you’re reducing that and providing the value back to your consumers. Uh, number two, within the revenue generation, we work with a lot of product management teams because we can very quickly create automations across domains, across technologies, it basically enables the product and strategy groups to create new products that they can go to market with, right? So, how do you do a combination of SDUN and security, creating the SASE product? Uh, how do we do a combination of wireline and wireless wireless services?

Karan Munalingal • 03:29

Right. So these are some of the things that we’re accelerating with the product management group. And finally, you know, reducing the time to go live with your 5G edge infrastructure investment. Right. This is where time is money. The faster you roll out your infrastructure, the faster you have it service ready, the faster your because consumers can take advantage of that. Now, on the other end, the way that we work with our service providers is around OPEX and operational efficiency.

Karan Munalingal • 03:57

Right. This is where we’re reducing, you know, all the Manual effort that potentially goes into managing and operating your entire network that support the revenue, right? How do we bring productivity? So you’ll notice numbers like, hey, you know, we reduce the time it takes to deploy a brand new network, whether it’s core, edge, from six months to one month. Now you have five months where you can actually capitalize and generate the revenue off of that infrastructure. If you look at all the potential outages that might have happened due to manual misconfiguration, uh, this is where, you know, from an ITential perspective, we bring a lot of validation and config compliance to the table, where our operators and engineers can put in their day zero, day one, as well as day two compliance strategies and policies to make sure your network is always up and running in order to support the services that you’re providing to your end consumer.

Karan Munalingal • 04:56

Right. So this is how we actually drive business value for Verizon as well as some of our other service providers. So let’s talk a little bit about the operating model for AI infrastructure. For the longest time, I think we have talked about, you know, a lot of our customers going from no automation to automation to orchestration. But now everyone is talking about how can they incorporate AI and take advantage of that reasoning on top of all the automations that they have already built, right? So if you if you’re looking at this slide right now, instrumentation is key. This is how we securely connect to anything in your environment a device, a controller, uh inventory system, an activation system.

Karan Munalingal • 05:40

I think when you have a normalized instrumentation layer that enables you to securely and at scale perform actions against your infrastructure, now you’re foundationally helping build that layer where you can automate and orchestrate on top. So this is where we bring in the deterministic uh capability from uh from an orchestration standpoint, right? So when you’re performing one task, writing a script, it’s a deterministic script, right? There is no uh if or else. This is where you know exactly what you’re gonna do in the network and you go do it. Same thing when you’re doing multiple coordinated activities, this is where determinism comes into picture. So from an ITNTR standpoint, the capabilities that we’re providing within the platform support deterministic automation and orchestration.

Karan Munalingal • 06:28

And we’ll talk a little bit about that during our demonstration and and future slide. And finally, as part of our final up uh the top stack of our operating model is AI reasoning. So for the longest time, we had a lot of our customers, including Verizon, build automations orchestration with a lot of business logic and if and else and fallouts. Now, Verizon has an opportunity to take advantage of AI intelligence paired with deterministic flows that take actions against the infrastructure. Right. So this is where Itential has introduced our Flow AI product portfolio in support of our customers being able to not only expose deterministic capabilities to agents, but build agents within Itential’s AI ecosystem. Right.

Karan Munalingal • 07:16

So this is how this is what we’re leading to market, working with a lot of service providers in making sure that their infrastructure comes along with their IT in support of the overall AI initiative. So let’s talk a little bit about instrumentation, right? Um, you see attentional in the middle. Our goal is how do we easily and securely connect to an uneven landscape of technologies that might have CLI, API, NetConf, et cetera. So, what you’re looking at right now is when you foundationally can connect and integrate and work with any technology, it basically widens the breadth of capabilities that your AI agents can have in the future, right? So, this is what we call our instrumentation layer, where we can very quickly, and we have done this effectively with Verizon as well. How we have uh very quickly integrated into your proprietary systems internally, commercial off-the-shelf systems provided by vendor.

Karan Munalingal • 08:13

At the same time, when there are new vendors coming with new automation capability via APIs, the goal is how do we quickly get them on board as API driven capabilities as well as tools for the AI world, right? So that is your instrumentation, that is your foundation on how you take actions against your infrastructure. The 2nd thing I wanted to quickly discuss is our portfolio evolution around the deterministic capability, right? So I’ve kind of talked a little bit about how do we uh take advantage of our connectivity, our integrations into your world. At the same time, how do we also improvise things like compliance and validation, right? That is some of the some of the very beginning stages when we work with Verizon. This is something that we did as part of your end-to-end flows for managing and maintaining your infrastructure.

Karan Munalingal • 09:08

If you look at the layer above that, we’re talking about service orchestration. This is where a lot of different teams would take advantage of our workflow, low-code capability to build out orchestrations across various different technologies and domain. Finally, uh the two new pieces that we have recently introduced is lifecycle management. This is where, you know, Itential has introduced the ability to remember and care for attributes about what you’re automating and orchestrating within your infrastructure. Before someone built out a workflow and it ran like a fire and forget workflow, right? All the data was saved somewhere. But now our customers have an opportunity to specifically remember certain things that might be coming back from the network, something that you were putting into the network in support of a resource such as a service, right?

Karan Munalingal • 10:01

So if you’re doing a layer two service or a SASI service, now we can remember what IP address was associated, which policy was uh actually created for that customer. So these are all the things that now become stateful, lending itself to a structured set of data, powering your AI and agentic experience up top. So me leading into our agentic and reasoning layer and how we’re actually going to be providing our customers the ability to create goal based or React agents that can now take advantage of. Validation, instrumentation, and statefulness, where you have your structured data in order to drive all the queries that are coming down from an NLP interface or from other agents who are trying to do intelligent orchestration based on reasoning. And finally, um, you know, self-serve, that has always been a thing from an itential standpoint. We want to make sure anyone, anywhere, can take advantage of services that are being built against the infrastructure. Uh, but at the same time, now we have the advent of AI agents coming in and requesting things from infra, right?

Karan Munalingal • 11:12

So now we also have the ability to expose the same network as a service capability to agents that might be consuming it going forward. So a little bit about Itential site a lifecycle manager. So this new capability that was added to the platform, it’s an application that enables our customers to define a resource. A resource could be anything, right? It could be an end-to-end service, a multi leg service, or it could be a multi-stage uh uh operation like software upgrade, right? If customers uh usually like to do software upgrades across different stages. They like to park their release, they like to push their release, they like to revert, etc.

Karan Munalingal • 11:53

So these could now be done in different stages. But as these actions are taken within the LCM application, Itential now remembers what it did. So it basically will give you a history as well as a diff of what is changing as part of that service or that device’s lifecycle, right? Hence, uh what we’re doing with a lot of our customers is working with their operations and engineering team to help define what a resource will be and what kind of actions are you gonna allow on that resource. So when Downton goes into his demo later on, he’s gonna talk a little bit about how we do that on 5G infrastructure and a end to end revenue generating service. So the 2nd part that we recently added was our Flow AI framework. So this uh AI product portfolio from Itential is to help support our customers and taking the next part of their journey from orchestration to an AI-enabled operation, right?

Karan Munalingal • 12:53

So one of the strategies on the left is the 1st is your integration. How do you integrate if you have an existing agentic or agent orchestration platform? How do you still take advantage of Itential as the infrastructure orchestrator and expose its capabilities safely to other agents that could be building uh intelligence up top? Right. So this is us enabling existing teams within Verizon that might have investment into Gemini that could now take advantage of Itentials. Um Infrastructure automation orchestration capability very safely but at scale.

Karan Munalingal • 13:31

So that way you’re not having your agent go rogue, right? This is where the key platform capabilities that itential provide come into picture around RBAC and auditing and eventing and logging. So we’re going to talk a little bit about that during Joksan’s demonstration as well. The 2nd way that a lot of our customers are thinking about this is within iTunchal itself, you now have access to every API, every project that you create, every workflow, every template. How do we now enable you to create an agent on top that could be an autonomous agent that could be doing things 24 7? Right. So now you have two different ways on how our customers can take the next leg of their journey into the AI world.

Karan Munalingal • 14:16

And we’ll discuss that in detail further. So uh let’s talk a little bit about our partnership with Verizon for the last eight years. Um, you know, all three of us, Dalton, uh, Joksan and I have had the benefit of working with various different teams at Verizon uh in order to drive a lot of different types of automation and orchestration projects uh that drive different various different types of value, right? So uh I’ll quickly talk a little bit about why uh Verizon actually chose Itential in the beginning, right? Service delivery acceleration was the number one reason, right? There are products that Verizon needs to drive to market, but how do we deliver it faster in a standardized fashion and with the best consumer experience? That’s where orchestration comes into picture.

Karan Munalingal • 15:04

Very deterministic, very integrated into your existing OSS stack, which drives further validation that we can do it faster, but in the same way for every customer. Uh assurance and compliance was a big thing in order to continue creating more product sets uh to go to market with. I think your network has to be stable and in compliance at all times. So, with respect to, you know, why Verizon initially chose Itential, our 1st initial project was upgrading core routers, right? This is where we actually got our teeth hooked in, where we orchestrated end to end how to upgrade core routers that run cities, right? This is where the assurance came in. This is where the compliance came in, right?

Karan Munalingal • 15:48

They trusted us in order to go do this at scale every night. And finally, it’s a robust integration into your existing ecosystem. I think one of the requirements from Verizon was we have a lot of proprietary systems internally that need to be part of our orchestrated journey. So our capability to very quickly integrate into any technology within Verizon ended up being one of the top reasons why we were chosen as a central orchestrator across these teams and across technologies. So let’s talk a bit a little bit about existing production deployments, right? We’re have been working with our transport and core teams in order to effectively drive automated device onboarding, config backups, changes across a different, a very different set of uh vendors, as you can see, right? It’s Arista, Cisco, Juniper, and Nokia.

Karan Munalingal • 16:42

Anything that has CLI API, we were able to very quickly integrate into and sometimes NetConf, right? From our standpoint, I think it was very critical that we also were able to adopt existing Verizon assets that you see called out at the bottom. Right. When we go into some of these teams and we went into some of these teams, we were able to operationalize their existing scripts and playbooks in order to drive an orchestrated onboarding process for the transport and core assets within Verizon. Number two, uh sell side backhaul, right? On the mobile mobility side, the mobile side, uh, we took advantage of the same capability, but now pointed that to the Sienna and the Cisco and the Nokia assets, right? So you think about how we were able to reuse assets that we have built with the transport and core.

Karan Munalingal • 17:32

And those were reused now for other vendors in a completely different domain. So from a time to market, time to value, where we were able to work with this new team and create these automation rather quickly because we had a lot of reusable Lego blocks that were available from the previous engagement. And finally, on the revenue generating side, you know, we have worked very closely with the MPN team internally to help drive end to end orchestration for enterprise customers to be onboarded for that service. Right. We did pre-checks, post checks, integration into existing uh OSS system to help make sure that we are keeping inventory in sync, uh, to make sure that we’re updating end users on when their uh service is being delivered, what the status of that service is. So that was very critical for us to do because we actually hit on a lot of different vendors in the 4G space and also during the migration in the 5G space, right? It’s from a Verizon standpoint.

Karan Munalingal • 18:34

I think when we got engaged, we started in the 4G C domain, but very quickly the past few years, we have built automation and orchestration to help every one of those customers migrate, start their migration into the 5G core uh in order to take advantage of the new capabilities in the 5G side. So the value realization, right? This is huge, both for us and Verizon. I think uh we’re very proud to work with the teams that we did uh at Verizon in order to drive this sort of value. So you’ll notice on the left in our early engagement with the infrastructure engineering team in 2017, we actually normalized and orchestrated software upgrades across the core router, as I mentioned. But if you look at the numbers, they’re mind blowing, right? Initially, they were used to they used to do this uh in a single maintenance window twice for two routers or one router when there is a failure.

Karan Munalingal • 19:28

But post automation and orchestration with all the checks that we put in, they were launching 80. Software upgrades against 80 core routers. And you’ll notice a time difference 30 minutes to two to 5 min , right? This is where you start hitting scale and how effectively you can manage all the assets within Verizon’s environment. Same thing we did with the CSR upgrades. You’ll notice the big difference was before it used to take 53 engineers to manage that entire project, came down to two engineers because they brought all their knowledge into the platform to build orchestration with deterministic logic. Number two is a consumer services.

Karan Munalingal • 20:10

So all we, you know, we had the opportunity to work with Verizon to onboard 17,000 enterprise customers for part of that service, which is a 4G MPN service, right? And now we’re in route to migrate majority of these to 5G in an orchestrated fashion. Not a lot of humans involved. But you’ll also notice the big saving of 510,000 cumulative days is because we went from doing this on a single customer from 30 days to two days effectively, as part of all the orchestrations that we have built. And finally, you know, on the mobile infrastructure side, the team that we work with saved 300,000 hours as they were bringing up new infrastructure into the mobile space, right? If you look at the backhaul, this is where they were able to leverage the existing automations and orchestration and plug them in against the backhaul vendors. And they were able to very quickly run 98,000 jobs.

Karan Munalingal • 21:12

you know, last year and then save 300,000 hours collectively. So this is just a subset of all of the value that we have driven at Verizon and we look forward to doing more. As the next step, I’m going to hand it off to Dalton, who is going to be doing and showcasing demonstration around our lifecycle management capability.

Dalton Smith • 21:31

Hello everyone. So I will jump into lifecycle manager and what we’re going to do. So we’re going to take on the um the the use case of moving from 4G services to 5G services. So what we’ve done and been working with the uh MPM team at Verizon, we’ve been uh we’ve identified a couple of the key business processes that you have to go through. One of the very 1st things you need to do is uh verify the data that’s available uh regarding the service for that customer and and verifying that the systems in place are ready to be migrated. The next step would be going through and staging the new network, the new components of that network because going from 4G to 5G is it’s a it’s somewhat similar, but there are some uh complexities where you will have to migrate uh certain components to new equipment. So you can imagine going from an LTE core uh 4G to uh 5G core, which is uh involving a UPF uh SMF.

Dalton Smith • 22:34

Um but these two steps that we’re talking about now one is uh verifying and validating, reconciling the uh information of what the custom where the customer is today, and then uh stepping into where you are uh staging the uh the change, the migration. Both of these steps could be uh handled during a business day, uh so it’s non in non-service impacting. And then the final step would uh take place during a um during a maintenance window where there’s a potential for slight impact with the customer, but you have the ability to be able to make the the changes um at a hopefully a less service affecting time frame. So there’s three main steps that we would go through again migrating, uh verifying uh migration readiness, staging, and then actually uh implementing. And as a final step that I think we all uh like to go through, we all like to go through, is this validating. So you uh before you start, you would uh go through a pre check and then uh again look at a post check, and then you would validate uh both the checks actually uh result in the uh results that you’re looking for. So that is the um kind of a the demo in a nutshell.

Dalton Smith • 23:52

So we’ll go ahead and jump into the actual uh demo itself. So, what you’re looking at right now is our uh lifecycle manager. And then on the screen right now, you’re seeing a customer, and we actually have we’ve named them customer A. What we are going to do is actually go through the process of one verifying the migration and their readiness. And then 2nd , we’ll go through a staging process, and then we’ll uh eventually execute and migrate this customer from 4G to 5G. Along the way, we also have other processes that could be implemented if needed, such as a rollback. So if you do encounter a situation where while you’re going through the migration, you do need a rollback, we have that action, which is something that um

Dalton Smith • 24:41

Karin had mentioned earlier in LCM. And again, these actions are business processes. We do have the ability for rollback. And we also have the ability to even decommission if we needed to, so we could remove that customer. So we do have various stages that could be our business processes that could be implemented. So just looking at the very 1st step, I’ll look at the properties and the parameters that we actually have. So what we have right now are what you’re seeing on the screen is just representation of the parameters, the variables that are associated with this customer, uh this customer service.

Dalton Smith • 25:19

And you can see that they are in a 4G state right now. They have been completed. The migration was completed, and basically we went from 3G to 4G. Um we also have information about the routing associated with the PE and the CPE of the uh of the of the network. So all of the parameters are set, and we would uh either place these in manually or we would actually get those from um Verizon systems. Now, as far as being able to validate, however, while you may have all this information, you still want to be able to go through a process where you’re reconciling and verifying that what you have in LCM and on the on the systems is reconciled and is correct. So we’ll go through the very 1st step, which again is just verifying migration readiness.

Dalton Smith • 26:13

So we’ll go through the step. There’s a workflow and automation in the background that’s actually running that actually goes through that that uh process. So we’ll we’ll look at the history, and we’ll switch over to history real quick. As the uh action is running, the automation in the background is running, and we’ll actually notice that the um operation is running and it takes us a few seconds, and we are actually uh in a position where we’re ready. So the reconciliation has happened. Um, and when we verified all the information that’s available today is it is as expected. So you can see the diffs, so our initial state, and then we um changed over to migration ready.

Dalton Smith • 26:56

So the very next step of the process, of course, would be to go through and we would stage the uh the the necessary equipment. So 1st we would enter in information now. Why why you are seeing um this in a manual format, this is for demonstration purposes. You can imagine that this could be handled through uh various means. So it could be an API call, the API call could be generated from an AI agent or uh various systems potentially. So what you’re seeing on the screen does not represent necessarily the the exact business process that Verizon would necessarily have to follow. So let’s go ahead and make the updates that are required, and we’re going to be making updates to the routing components, and we’re also going to be choosing a new uh a new router or new CPE that we’re going to be working with.

Dalton Smith • 27:49

Instead of working with CPE one, we’ll be working with CPE two. So we’ll go ahead and go through the process of staging the um the equipment and uh the the records as well and changing the um the state of the record for the serve for the actual service this one customer from a um from migration ready to stage like or migration uh staged. Now let’s switch over to history and again this process is running in the background. We’re actually going out to the systems and making the updates to the routers and uh also updating our records and uh in a few seconds we’ll actually see the update um take place and we’ll be able to see in our history through LCM the two different changes or the two different states that we uh were working with. So Okay, so we actually went from migration ready, where we actually hadn’t changed anything to uh migration stage. So we have updated all of the necessary records uh associated with this one customer and are uh in a position where we’re ready for the very next step, which again would be taking what you’ve actually staged and executing or applying that migration.

Dalton Smith • 29:18

Again, but the 1st two steps you can imagine doing this during a normal business day. The final step would be going through and um actually going through and uh executing the migration itself and applying the migration. So let’s go ahead and we’ll start that process. And as a final spell safe, again, we uh would be going through the process of um Entering information again, this could be driven by an API call or uh various other means. So this isn’t something that’s required in order to be able to do this, but it’s it is uh for demonstration purposes a manual step. So we’ll go ahead and kick this off now.

Dalton Smith • 30:00

Unlike all the other um actions, I’m going to drill down into the actual workflow itself so you can kind of get a sense of what’s actually happening. And we’re looking at our uh job queue. And we’re getting ready to look at our workflow that we actually have running. And drilling into this, you can see that as it’s going through the different steps of the process. So one of the very 1st steps that we go through is again performing a pre-check. So we’re actually validating and verifying all the different systems are ready for the migration before we actually even touch the systems. The next steps are actually going through and updating the PE and the CPE associated with this customer, and we’re shutting the parts of the network that are actually associated with the change in the migration.

Dalton Smith • 30:53

As we’re going through this, we’re again recording all this information and making this available at the end of the of the migration itself in LCM. And you can see that we’re actually shutting down. So we’re we’re shutting the actual network. And then we’re now we’re bringing up the new network or the new components associated with this service, this 5G service. And then as a final step, and again, this is a manual step that you can actually see inside of ITential, we’re actually presented with information about that post check. And you can see that everything passed. We have migrated the customer from 4G to 5G, and uh the customer is uh hopefully happy at this point.

Dalton Smith • 31:53

Now Now, switching back over to lifecycle manager, we’ll look at the history just to kind of give you an idea of uh sh I needed to hit the success so we can close out that one automation. And then we actually are going to be in a position where we should be able to see the historical information related to the the actual um Migration. So we went from a migration stage to a migration complete. And then we also updated the actual information related to the uh routing and the customer itself. So we not only went from the 100.2 network or 100 network to uh the 200 network, we also updated the records to reflect that the uh customer now is running five uh running in a private 5G uh type service.

Dalton Smith • 32:45

So again, just thinking through this whole process is as you can see, we’re actually watching and keeping track of all of the uh information. And what we’re doing, we’re not only um relying on the sources of truth, we’re also relying on or looking at a um source of confidence. We’re actually building a source of confidence that we can actually kind of work with. So this is something that’s very important when you start thinking about the structured data that we’re working with, as well as uh that that one statement. And this is something that will be very important when we start looking at uh working with some of the AI uh components that we actually are going to transition to next. So thank you. Joksan.

Joksan Flores • 33:32

Awesome. Thank you, Dalton, for that demonstration. Um, now we’re going to be focused on transition a little bit, talking about the agentic solutions that we have to offer with a newly released uh framework that we’re working with. And one of the components that this framework includes is the flow MCP capability on top of Ittential Platform. Um, so one of the pieces that we’re bringing up, in addition to what Karn had mentioned earlier with Agent Builder and the agents that live on the platform, is the Flow MCP. And what that gives us access to is we can expose Platform capabilities as tools to agents that are built inside of Verizon.

Joksan Flores • 34:11

Karin mentioned some of the initiatives that are going in internally using Gemini and so forth with the internal assistance. And the whole idea here is to expose various sets of capability that Ittential has to offer directly to those assistants for exposure of services. Dalton kind of explained and alluded to this a little bit, right? Some of the things that he was doing regarding triggering actions and executions of stateful services on the Ittential platform. He was doing it all manually for demonstration purposes. But these are some of the things where we can start leveraging agentic reasoning and getting value out of those systems by plugging them in into Attential. And this is virtually a ready to go thing, right?

Joksan Flores • 34:53

For today’s demonstration, I have designed a very simple agent that has access to a certain set of ital tools. And it has the capability of doing a bunch of things like executing workflows, reading data from service instances in lifecycle manager, reason through them and creating some reports and things like that. Now, kind of continuing with the whole theme of the demo, we’re going to the next slide and we’ll see some of the things that we’re gonna focus on regarding um validating the service execution, validating some of the health of the devices that Dalton kind of has touched on. So when you start looking at driving the lifecycle of a change, right? We kind of did some day two activities type of things, right? So we took a service from 4G to 5G, moved the routing over, migrated from an old CPE to a new CPE, moved the routing layer and the routing policies along the way with it. But there is a bunch of things that we have to do after, right?

Joksan Flores • 35:52

Dalton did a little bit of validation by just using some show command execution on the CLI devices and so forth. But there’s also a bunch of stuff that we have to do after, right? This is meant to represent here some of the things that we normally would do, right? We would make configuration changes to a device after a service turnout. We’re not gonna focus on that. Dalton did that already. Then we would normally go and do the device health check, making sure that routing is healthy, memory consumption, processes, CPU, all that stuff is good after we make a traditional standard change in the device.

Joksan Flores • 36:22

Then after we validate, we don’t have to do anything else drastic, like performing a rollback or anything like that. We want to go ahead and establish a backup procedure, right? Well known, good configuration. We’re going to go and execute a backup. Normally we could go after the fact, do config compliance for remediation if we have to. Some config compliance violations. This process, we’re going to skip it.

Joksan Flores • 36:43

Dalton has a well-designed state flow service that he provisioned on. Those are very much validated, deterministic flows that provision config that is well known and tested in the ITENTL platform. So we we’re going to go ahead and go with them, bypass that process. And then we also want to do some software version compliance. So that’s kind of one of the demo. And then we’re going to do some validation and look at some cool things on some of the stuff that Dalton has done. So we’re just going to go kind of continue on with that.

Joksan Flores • 37:17

Okay, awesome. My crew, my screen looks okay there. Yes, sir. Perfect. Thank you. So, what we have done here is we have an agent design in the background that has a certain series of tools from Ittentional Connected as well as its personality, right? Just very simple personality that says, hey, you’re an expert on Attentional platform and you have access to certain uh certain tools.

Joksan Flores • 37:40

Um, just kind of dialing um double clicking here in the background a little bit. I have onboarded the attentional MCP and I have provided a series of tools, right? So you can see tools here from running command to start workflow, getting instances and so forth, as well as some health check capability. So by virtue of onboarding MCP and enabling those tools, which a lot of them are provided by deep by default by MCP, or some of them are just workflows that we design in Itential, I have now enriched my agent to provide it a series of capabilities that it normally wouldn’t have, or I would have to code my way through. So what we’re kind of trying to say here is Verizon has already consumed Itensal for a long time. There’s a lot of capability on premise today that can be used and leveraged by just by virtual or just onboarding something like Attentional MCP and exposing it to your agents. By starting my session here, I said, hey, hello, summarize your capability in a few bullet points.

Joksan Flores • 38:32

It says I’m Claude connected to Attention Platform. I got capability for workflows, managing devices, managing lifecycle, monitoring jobs, and executing compliance and health checks. Right? So that’s one of the things that we want to focus on. So I’m going to start typing in some prompts and then we’ll start working through some of these um some of these requests and see you know what things we can kind of glean and reason through from uh the data provided on the platform. So I want to say, can you get all the lifecycle manager resources and find customer A on the migration customer migration service, which is what Dalton was using, and then provide me a summary history for this customer. So, what I’m trying to do here is I’m trying to kind of drill down on some of the stuff that Dalton was showing earlier.

Joksan Flores • 39:23

And this will take a 2nd . Um, but essentially what we’re doing is trying to drill down on some of the history items that Dalton was showing along the way and some of those diffs. And what we’ve done is we’ve exposed that capability via MCP to the agent, and the agent will now reason through all that data and provide us a natural language response of what happened. Right. And that’s kind of what it’s done here. So now the agent has done its job very quickly while I was talking, and now it’s provided us, you know, it kind of executed a handful of tool calls, right? The get resources.

Joksan Flores • 39:55

So that means it went and retrieved the items to lifecycle manager resources. It got the instances for the customer migration service, as we indicated, and it got the action executions, which represent the history for that customer. So it’ll say, I found the customer A, and here’s a comprehensive summary of the history. So this is extremely cool because we can just now Glean data from what happened to that customer without having to necessarily go through a series of JSONs, right? What Dalton was shown was very demoy, right? Things that we want to do in the background.

Joksan Flores • 40:26

And now we can actually leverage AI reasoning on something like a chat bot tool, or perhaps pull this information and post it as an HTML on a website internally to say, hey, this is what happened with that service. So this is my the service state is migration complete. Service type is 5G, class is bronze, and the current network configuration is now PE node on this AS and this IP and device is CPE2. Prior to the transition was actually CPU CPU CPE1. So it’ll say here it has gone through some trans some uh lifecycle actions, customers completed a few migrations that we’ve done for testing, right? There’s a bunch of those because we tested some of this stuff. Um it’s gone through these key migration phrases, verify migration stage, execute and some testing, rollback, and all the stuff that’s kind of gone, right?

Joksan Flores • 41:14

Gotten through. So there’s a lot of data here that we can glean from just that service and the history um that has services gone through. And like I said earlier, um, after we kind of grab history and reason through some of the data that has been provided by Life Cycle Manager and some of the things that that service has gone through, we actually have want to go and enhance um and tap on on that validation, right? That last stage that Dalton was talking about. So, what I want to do is the 1st thing that I want to do is I want to execute a health check on those devices. Uh, can you execute a health check? On devices um CPE 1, CPE2, and PE East.

Joksan Flores • 41:56

So those are all the devices that were involved in the migration. CPE1 probably less important now because we have migrated from CPE1 to CPE2. But perhaps I have other customers that are hosted off of there. So maybe I just want to keep a history on those. And then I want to evaluate the rest of the other ones. So I’m going to go ahead and submit that. This will actually take a 2nd .

Joksan Flores • 42:16

So I want to see what it’s actually happening. Okay, so we’ve seen that now after submitting this, it’s gone and executed a device health check call, which happens to be a workflow. So the agent has now been instructed after a workflow is executed, it’ll track the job ID and it’ll actually wait for the job ID to be executed. And like you’ve seen, we got a lot more details here. Some of the stuff can be summarized, but it’s kind of nice to see under the cover a little bit, right? So the way that the health check works is we actually spawn a bunch of mini um child jobs, right? A bunch of mini modulo automations that are each executing a particular check against the infrastructure.

Joksan Flores • 43:00

So in this case, it happens to be you know a handful of those 21 of those. Those can be optimized and reduced, but in this case, that’s kind of how we design it. So now that’s uh going on. So I’m just gonna wait for that to happen. Okay. So the job has completed successfully. Now, one of the other things that this agent has been instructed to do, and just because it’s very important, right?

Joksan Flores • 43:33

A lot of times we, you know, we like the starting with an AI strategy. The very initial phase is typically going and having some sort of chat assistant that we can talk to and ask it questions and so on and so forth. But I think eventually we will need AI to help us kind of augment the way that we do certain things today. So one of the things that we have instructed this agent to do will be to actually create a health check report out of the data that it has received. So this will take a 2nd , but it’s actually going to go ahead and do that, or it’s going to go ahead and create a health check report with all the data that we have um that we have collected from the environment. So I’m going to go ahead and wait for it to do that, and then we’re going to go ahead and observe and kind of read through it. Okay.

Joksan Flores • 44:33

So our report has been created and it’s uh being rendered. Now look at that. Um, and it’s got a lot of uh very cool colors and very cool sections in here. So we have a revive device health check report. We got an accessory summary, summary of all the checks and evaluated 21 distinct health parameters across all devices, resources, interface, status, connectivity, and operational metrics. The devices appear to be healthy. Um, you know, we got some some reasoning here through some of the meaning of the devices uh naming convention and things like that.

Joksan Flores • 45:05

And now more importantly, I get details of all the health checks, right? So we check BGP, we check interfaces, we check VRF, we check CPU load, connectivity, and so on. And then we also get some things here, some recommendations as well. Notice that while we’re all along doing this, I have not logged into the platform once. I have the platform open in the background, so I could technically flash over there and look at it, but I have not logged into it on purpose because I want to interact with it with this agentic strategy, right? Make sure that my chat and if my chat agent has all that data that I need and provides me all the information that I want from it. The next thing that we’re gonna do is we’re gonna go ahead and execute device backups and software compliance.

Joksan Flores • 45:49

For those devices. So I’m gonna go ahead and do that and let’s see. This might take also a little a little bit, but we’ll see how long it takes. Um, but now after my request, it’s already gonna go ahead and trigger the device backup workflow and see what it does with the software compliance. If we we have to wait or not. But typically this is something that you would just do launch, oh, yeah, it’ll launch them both in parallel. That’s fantastic.

Joksan Flores • 46:12

That’s kind of what we want. So it launched both workflows in parallel after it notice one or the other, it’s tracking both job IDs. So now it’s gonna sit down and wait um how much time has been elapsed. And it’s gonna go ahead and wait for okay, so the device backup completed. Now it’s gonna check on the software compliance status. Okay, now it’s going to go ahead and actually update or create a comprehensive report. So now we’re going to get backup on compliance results.

Joksan Flores • 46:43

So let’s wait for that to happen and see what type of thing, what type of metrics we will get. And the reason why we kind of spent some of the energy on instructing the LLM to create HTML reports and things like that is because this is the type of thing that we’d have to normally report to either change management or through our executives, right? They will want to know a summary of what things happened, right? What customers were migrated, is the network in a healthy state, things like that, right? All those things. So this is where this becomes extremely useful, right? We still have the attention platform in the background doing all the heavy lifting, executing all the work, executing all the compliance, executing all the health checks, making sure that the CLI and the control plane and the data plane outputs look good.

Joksan Flores • 47:27

But then we utilize AI reasoning in order to create these pretty dashboards and kind of reason to some of that data. So now our report has been updated. We got backup and compliance complete. We got the summary. It’s been successfully completed. It took 17 seconds apparently and 12 seconds to do the software compliance report. And now we got a very much comprehensive report here for device backup and compliance as well.

Joksan Flores • 48:09

And saying all the operations that were completed successfully. So we complete it. After we validated that they can fix, we’re good on the devices. We don’t have to hopefully roll back the service or anything like that. Now we got backups of what they can fix. We got compliance verified on all the devices. And then we got all the details about this, right?

Joksan Flores • 48:26

The jobs, how long it took, everything that was verified with some checklists for the devices, backup success, compliant. Everything is complete and verified on the devices. And we got the backup location to be set to a secure repository. Mind you, we use workflows for a lot of these things. So the backups could be stored anywhere. This is one of those big use cases for Verizon, right? Performing backups at scale.

Joksan Flores • 48:47

So this is one of those things that we consider to be super important, um, as well as validating that the software is in a good state. And the security patch and the level where we’re putting our customer, we’re moving our customer services is on a certified software version. So we go through all of these and then we get a summary recommendation on next steps. So now out of this, we have two artifacts. We have the device health check and device pack of compliance reports that we can provide to our management and to our change board to say, hey, all the changes that we executed on the environment are certified and everything is good and we validated with all the checks that we had in store. So thank you. This is a um very, very good result.

Joksan Flores • 49:28

And uh back to Karan.

Karan Munalingal • 49:32

Thank you both uh Dalton and Joksan for wonderful demonstrations that highlight Itential’s you know stateful orchestration capability as well as agentic uh infrastructure operations, right? They go hand in hand. I think um what Joksan kind of uh dovetailed on, you know, Dought’s demonstration on leveraging existing data and actually creating a report for an executive or engineer. I think that was very critical on how all the uh pieces of the puzzle come together. So, with respect to how Itential is actually defining the future of AI driven automation, as everyone at Verizon who has worked with us know, right? Orchestration and automation is key to driving this forward. But not only that, when you take a platform approach, all the key capabilities that come within Itential, uh within Itential’s platform around.

Karan Munalingal • 50:25

You know, um, R back, authentication, eventing, integration. Like these are all the core capabilities that are going to drive the AI ecosystem forward more securely, especially against the infrastructure. Right. The other thing is how do you pair determinism with reasoning? I think for the longest time, everyone has now gotten comfortable with dragging and dropping and writing scripts that do very deterministic action, but we have also incorporated a lot of business logic. Can we now take full advantage of Itential’s integration into your LLM stack and AI reasoning in order to make it more efficient, more intelligent, and more abstracted? And finally, if you talk a little bit about an open and extensible enterprise grade platform, this is where the instrumentation comes into picture.

Karan Munalingal • 51:18

Verizon will continue to invest in a lot of new vendors in the market that drive their value in different domains. How do we bring all those vendors and their capability to bear as part of your orchestration and AI for infrastructure strategy? So, you know, that was the end of our presentation. Uh, given time, we do have a few questions that have popped up in our chat. I’m gonna be moderating some of these. Um The 1st one being, you know, let’s go straight to the AI by John.

Karan Munalingal • 51:52

Where is the AI getting the data pool to answer your question? So I don’t know, Joksan, if you want to uh uh take a stab at this question around your demonstration and how you know the NLP interface, how it interacted, where it got the data.

Joksan Flores • 52:06

Yeah, I um I replied to John in the chat, but for the benefit of everybody else, right? So the way that we have it set up is um there was one of the slides earlier where we were talking about the MCP layer being on top of the platform and allow us allowing us to expose capability to the AA models. That’s how it’s working. So we have configured an MCP layer, right? An MCP server that connects to the item to platform via the Northbound interface, and that MCP has a curated set of tools that provides the AI capability to access data from the platform, like executing workflows, getting workflow outputs, workflow job results. Going into lifecycle manager and pulling instance data, et cetera. So, John, hopefully that helps um answer that question.

Karan Munalingal • 52:49

And Joksan, one thing I would add to that is Verizon and every other customers that have invested in our platform have full control of what an agent could do through the platform, right? I think data is starting to become gold. So when you talk about what northbound asset actually has access to Verizon data, this is where teams within Verizon have full control over which capability could be invoked, which data set could be exposed for any level of reasoning or reporting.

Joksan Flores • 53:20

100%. Yeah. If you looked at my demo, I kind of glanced at the capabilities very quickly. I exposed probably 10 for the purpose of this demo. We have over 70 tools, right? So you get to control. And on top of that, you get the R back capability of the platform as well.

Joksan Flores • 53:32

So

Karan Munalingal • 53:33

Perfect. Thanks, Joksan. Um, the 2nd one that’s in the public channel is does LCM help with service provisioning use cases only, or it can be applied to other network maintenance activities such as upgrades and other orchestrated configuration management. So I don’t adapt to the case. Absolutely.

Dalton Smith • 53:51

Absolutely. So LCM um is not just focused solely on service provisioning uh use cases. While we uh kind of went through that uh somewhat, we uh you saw during the demonstration where we actually migrated. Um so we actually were making changes to um to the systems themselves. So you can imagine us uh making uh changes to um Again, the uh the configuration of the devices, you can also even imagine migrating from one type of system to another type of system. Um, while this may be service impacting, it’s not just solely uh service uh focused.

Dalton Smith • 54:29

So LCM definitely could uh address many different use cases. And we actually are uh working with many customers uh using LCM in a lot of different uh scenarios as it relates to uh to to uh being able to utilize uh LCM.

Karan Munalingal • 54:43

So and one thing to add, right, uh having experience with some of the other customers is we’re starting to see a lot of trend into any activity that needs to happen across different teams or across different stages, very distinctive stages, like you know, upgrading software across your infrastructure devices, uh lifecycle certificate, you know, management across this uh from a security standpoint, blocking IP addresses, et cetera, across your entire infrastructure asset, right? So I think these are some of the use cases that have become prime beyond you know, product and service provisioning for things like you know, MPN or layer two VPN, et cetera.

Dalton Smith • 55:24

Yeah, and and I think one thing also to really kind of emphasize when you start looking at LCM and what we bring to the uh table with LCM. Now is being able to structure the data that’s actually being used and being able to use that data logically and methodically through uh different business processes. Again, uh during my demo, we actually were utilizing the data that was available to make the changes and actually make the adjustments as necessary. So we also were tracking the state of uh individual service, but you can imagine the same thing with uh devices or even systems um throughout. And that structured data is very, very important when you start thinking about um having that source of confidence that uh would be then used by the uh AI agents and some of the uh the different things we’re actually uh doing today.

Karan Munalingal • 56:17

So awesome. And we have time for one last one. Uh this is a very common question. Um, what are some of the challenges that you have seen? And we’ll make it public. Uh, what are some of the challenges that you have seen related to adoption of AI into your business processes and automation practices? And Joksan, if you wanna take a crack at this, right?

Karan Munalingal • 56:41

Working with some of the other customers, what we have heard.

Joksan Flores • 56:44

Yeah, I think we have um there’s there’s a few things, right? I think the biggest thing to me is a very um, it’s a misunderstanding of the technology, right? I think um a lot of customers come to us and say, you know, what can your AI do? Is it gonna do magic for me? Am I gonna be able to create a workflow with just one task that will say, hey, go on configure this in the environment and let it. Go ham, right? And kind of do all sorts of crazy things.

Joksan Flores • 57:10

Right. I think this has helped us quite a bit kind of solidify our pillars, right? Karin talked about how we focused and this we’ve kind of grown into this model from what we’ve doing for a long time, from our capability on providing instrumentation, very solid integration into CLI APIs and so forth, as well as our workflows that remain deterministic, right? There’s a combination, right? I think this is the biggest challenge is to educate everybody to say you will need a deterministic portion of the environment that not only saves you money, right? Because it’s not consuming AI tokens, but at the same time, it’s executing the logic that you need every single time in the same deterministic order, right? The stuff that Dalton was doing today, I’m very definitely comfortable triggering those from AI in that order, right?

Joksan Flores • 58:00

That process that Dalton did, I can automate with AI very comfortably. I would not trust the AI to come up with a mop by itself to go and migrate a customer to from 4G to 5G and route the traffic appropriately and perform the validations that it wants, right? I think you know that’s one of the things that we spent a lot of time is educating customers to say there will be simplicity on the way that you’re gonna build workflows moving forward, right? Potentially less data mapping, less reasoning, less human intervention between stages and things like that. But you will still need the determinism of the platform. And uh, you know, hopefully that helps. And Karin, if you have anything to add, Dalton as well.

Karan Munalingal • 58:37

Yeah, I think one thing to add there, Joxton, to that question is I think trust factor uh is critical, right? As you mentioned, I think everyone is now comfortable with some level of automation, but now here comes AI that reasons with itself, right? So I think you know, the the whole concepts around human in the loop, human on the loop. Like these are going to now become critical for the next few years. So everybody within the service provider and enterprise space feel more comfortable allowing agents to do a read-only activity, sometimes write. So this is where the control and the security comes into picture is how much are you going to let the agent do it? That’s assurance.

Karan Munalingal • 59:16

Number two, that we have obviously all three of us have heard is culture. I think people have moved mountains to get operations engineering, everyone to go from no automation to automation to orchestration, and now incorporation of AI is the next challenge. So I think these are some of the common challenges across every one of our customers. Uh, but I think having a game plan on how do you slowly incorporate AI capabilities within your day to day gives you the confidence that you can do more with it. So I think that’s something that you know we have worked with a lot of customers. Would love to do that with Verizon as well, in order to drive more impact on how reasoning and determinism can and drive within the infrastructure.

Karan Munalingal • 01:00:00

100%. We’ll be here.

Dalton Smith • 01:00:02

Yes, absolutely.

Karan Munalingal • 01:00:03

So with that, uh, that was the end of our webinar presentation. So again, thanks to my peers, Joksan and Dalton, for joining me, as well as all the participants who hung out with us, uh, listening to us and watching through the demonstration. Hopefully, this was helpful and how we are working with other customers, but purely what we have done with Verizon over the last eight years and what we want to go uh going forward, do going forward in order to support your 2026 objectives related to you know OpEx transformation, revenue generation, as well as uh AI initiatives within Verizon. So thank you all. Um, have a great day.

Joksan Flores • 01:00:48

Thank you. Bye, everybody. Thank you.