Richard Thurston, TelecomTV (00:02):
We're here with Matt Whitman to talk about adaptive inventory. Matt, what have you got for us today?

Matt Whitman, Ericsson (00:07):
So we're talking today about unified inventory as a prerequisite for the autonomous networks journey and achieving higher levels of autonomous networks like AN Level 4, we're going to be showing the digital twin as part of the inventory management solution functionality within adaptive inventory and how we're enabling predictive optimization of the network. So when we look at building blocks for what's necessary to achieve higher levels of autonomous networks, real-time inventory and inventory that is connected a hundred percent accurate to your network is one of those building blocks that must be there in order to achieve those higher levels of optimization. We have a federation layer within adaptive inventory that allows us to connect to any EMS, any NMS as well as any other OSS system to take your planned view of your network and reconcile it against the as built view of the network to have that unified real-time inventory capability.

(00:59):
In our presentation today, we're going to be looking at the cognitive loop capability and the cognitive agents that run in that agentic AI and the digital twin will be working as the Proposal Agent and the Evaluation Agent in this scenario that we'll be looking at. So if we imagine ourselves as a planning engineer, a planning engineer's job is obviously to look at issues that have been identified in the network and things that need to action and or resolve. And when we talk about the autonomous networks journey, we're really talking about achieving and crossing that chasm between human-led machine assisted processes to machine-led human assisted processes. So if we look at our cognitive loop dashboard within our cognitive loop ecosystem, we'll see that there has been agents that have collected FM and PM events and we've identified an area of the network that requires some review.

(01:56):
So when we look at that particular area, we'll see a collection of FM or fault management and performance management events that indicate that there's a problem. What we'll do is we'll launch Adaptive Inventory's digital twin and we'll look at the results of the predictive agent and what it has identified as potential resolutions already in the background that have been executed in the digital twin as a scenario. So we have several options that have been identified in terms of providing relief for the congestion and the problems on the network. One of those of course, is to scale the access network function or the RAN function. It will add a radio to the network and then measure the increased capacity. Now, it will of course, identify that transport in terms of a bottleneck and a resolution will not be met with this solution. It also provides a cost estimate in terms of a low, medium, or high in terms of what the impact of the actual cost for the solution will be.

(02:56):
So here we're seeing that this is not really a viable outcome. The other option, or the second option would be to scale vertically the functions at the location. So not just a radio, but adding an expansion to the layer three VPN on the transport to offload the user plane traffic. Now this is a great solution that is going to definitely resolve the problem in the short term. However, in the background in the knowledge plane, you also have AI agents that are creating forecasts, and the Central London growth forecast indicates that the traffic expansion in the area in the longer term will outlive [grow beyond] that capacity. So although this is a cost-effective solution, it will not resolve long-term the problem or the issue in terms of what's happening in this particular area. So here we actually see the flag that says "Best". So the Evaluation Agent that's running on the product has already selected this as the most viable option and the best option in terms of resolving the problem.

(03:58):
So here we're going to see the deployment of a new gNodeB in the vicinity of the area to provide enhanced coverage and capacity. We're going to see that we need to larger, we need to have a larger expansion of the donor cells transport to offload that user plane traffic, and we need a microwave link from the donor cell to the actual new cell that we're going to deploy. And of course, in the scenario analysis from the digital twin, we see that the growth forecast, this solution itself will provide enough capacity to outlive that forecast for 30 plus months. Now, of course, this option is an expensive option from a CapEx and an OpEx perspective because it will be a much larger increase on the network. However, we are waiting at this point, those agentic AIs that are running in the background, they're waiting for a human to say, what selection should we do?

(04:52):
So that planning engineer's role here, is not to do the analysis and do the design and run the scenarios and figure out what's best to do. They are going to take the results of the agentic AI and make that decision. So here we've transitioned from human-led to machine-led. So what we can do is then proceed and the agentic AI has already identified the work orders that need to be created and run through the system in order to then execute this change. So we'll see that there'll be a pop-up here authorizing the work orders, and this is the only human interaction in this process. So we will confirm that that work order has been authorized and the system will start working. So now of course, obviously in the background, some point later, these network build work orders and change orders have all completed. And of course, the agentic AI cognitive loop that is constantly running the background, once these work orders have been completed, we'll see on the agentic AI dashboard that now there is no FM and PM events the future because the solution to those issues have been resolved and the congestion is no longer there. So what we're seeing is the value of the agentic AI and what it can proactively do to automate and really reduce the amount of time to have a positive change on the decision making process to address network issues and network conditions that the assurance solution is identifying or the assurance AI is identifying.

Richard Thurston, TelecomTV (06:26):
That's really powerful, Matt. Thank you.

Matt Whitman, Ericsson (06:27):
Great. Thank you very much. It was great to talk to you today.