How will we get to a real edge-computing-and-real-time future? That’s not an academic question; everyone in the tech space is waiting breathlessly for a rush of new opportunity. I think that most probably agree with me, that such a thing will have to be driven by new real-time applications. That’s also what enterprises think (which is why I think it’s the case), so it’s reasonable to ask how enterprises think we’ll get to this happy goal, if we do get there. Here’s what I hear, focusing on two groups.
In the last year, my first group of 181 enterprises told me they had some real-time edge-computing applications. All of these were based on local hosting, meaning on process control systems located in the facilities where the processes were sited. Factories (124) were dominant in the majority of the facilities, with warehouses/yards second (44), and the remainder divided among more specialized things. Of the 181, none said they were planning to take these applications to WAN-connected remote hosting in 2026, and so the didn’t see themselves needing real-time specialization in WAN QoS.
You’d think things were hopeless from all of this, but of the 181, 48 said that they believed they would “likely” eventually want to look at remote hosting of real-time components, and thus would be shopping for the best connectivity solution. It’s this group we need to focus on, because their views represent the first phase of any shift in technology—willingness to change things.
What’s interesting about this group is that 31 of 48 are in the warehouse/yards space and the remainder have at least one plant that spreads out over a significant area. All have at least one case where multiple real-time-equipped facilities exist within the same city. These 48 say that the way they’ll get to remote hosting of real-time components is by spreading out existing applications across facility boundaries. In every single case, the specific missions being considered involve movement of goods or materials, not production/manufacturing, and not direct worker empowerment.
The key requirement driving a chance of remote hosting is a requirement to coordinate the movement of things based on real-time considerations. Further, the group seems to see the application focus being on automated movement. Those 18 who commented further said that if something was being moved by human agency, the delays would be such that “real-time” coordination would not be useful; it would be what I’ve called a “transactional” drive not an “event” drive. This means that some sort of robotic/automatic movement would be involved, which means that the devices would need to be acquired and the facilities prepared for them. This would not be limited to the “movers” of the process; you’d need to be able to get things on and off those gadgets in an automated way or you’d still have a more transactional process. This group, then, represents enterprises who are replacing workers, not empowering them.
The second group of 27 enterprises belong in the utilities, refining, and healthcare verticals, and all of them have already initiated measures to empower workers. Not only that, all of these enterprises are focusing on non-office types, people who move around in their jobs, and usually move a lot. In 22 of the cases, the instrument of choice is a tablet, in the others a smartphone, and all the target workers fall into the top two quintiles in terms of unit value of labor.
Today, this second group does (staying with our past terminology) “transactional” empowerment, meaning that the device is used to deliver (and sometimes obtain) information that’s not necessarily real-time. Doing things like giving a doctor the ability to view radiology, a power engineer the ability to review the schematic of a panel, and so forth, is the goal. The future is perceived as spreading out in two ways—informationally and positionally.
The informational piece is seen as a primary direction by 15 of the group, who have already identified some real-time goals that would augment the information delivery requirements. But even this group admits that without at least some broadening of the area to be covered, there would not likely be a justification for real-time-capable network services.
The difference between the future and the present, for this group, seems predominantly video, but sometimes other telemetry that the preferred device can support would also be valuable. This activity, particularly where some form of real-time video analysis is projected, is interesting because it offers a potential pathway to a variety of applications where AI used in real time to analyze a scene could aid workers in some important task. I’ve heard examples in public safety and medicine, and there are also obviously military examples.
So this group represents the most generally valuable opportunity, and the application’s need to grow out of a transactional model presents both positives and negatives. One positive is that there are a lot of transactional applications that already empower workers. Another is that this application targets the workers whose productivity is most valuable, which is where you’d want to start a real-time-empowerment trial. The big negative is that there’s a significant gulf between real-time or event-driven empowerment and transactional empowerment. It’s likely there is no universally applicable bridge, so each worker/application combination would have to be examined to determine how to advance.
Eleven of this group, fortunately, had the same notion of the next step. You give the target a glasses-camera, body camera, or whatever works for the specific situation, and you feed video at appropriate times (which might be either by default or only on request) and you push the video through an AI analysis tool to look for some set of conditions.
There two things of critical importance here. First, nothing is going to create a telco differentiated-services revenue stream except stuff that has QoS requirements too stringent to support with a best-efforts service. That, to me, means real-time. Second, the service does no good until you can put the application in place, and the most prudent way for enterprises to do that is to evolve from the type of application that’s closest to what you want. They’re starting to see that application class, but support and encouragement from vendors and the media, aimed at the migration and not just the hoped-for end game, would help a lot.
