What’s more important to the future of tech, digital twins or AI? If you have to pick only one thing, I’d argue that the digital twin would win, because the future of tech depends on more real-time, real-world, automation and lifestyle augmentation. Digital twins that model the real world seem inescapable in that mission, and in fact are already being adopted there. But AI can also play a role, and as I noted in an earlier blog HERE, companies like NVIDIA are seeing the link, and the connection is further noted HERE. I think the case for symbiosis is clear, surely clearer than the progress, and sadly the article I reference doesn’t offer us much insight. Let’s try to get some from what enterprises tell me.
The article I cited above is interesting to me because, since it represents a survey and I don’t do formal surveys, it provides at least a slant on intentions. I rely on anecdotal comments from enterprises, which are hard to frame intent from reliably, but they do offer a view of actual use/implementation. However, my data is best used by considering missions rather than the vertical an enterprise represents.
Manufacturing automation is the place where digital twins are used the most today; 69% of enterprises involved in manufacturing say they are using digital twins and 100% are considering the use. Facility automation missions(smart cities, buildings, campuses, etc.) uses digital twins in 34% of all enterprises, and 73% of those involved in warehousing, utilities, refining and fuel/power distribution, etc. Beyond these missions, penetration of digital twin technology is limited; only 9% of enterprises cited other uses to me.
Right now, less than 10% of enterprises use AI in IoT-real-world-related missions at all, but the comments I get suggest that those who use digital twins are more likely to use AI, though symbiotic usage is still limited. A big part of the explanation is simply the novelty of the AI and digital twins individually, and the fact that any relationship has been promoted only recently (in 2025, in fact).
There are three ways that AI and digital twins could form a symbiosis. First, they could be used for “adjacent” elements of a real-world-automating process, meaning they cooperate to do separate and discrete things. Second, they could share responsibility for a step, with a flow of information between them being essential to the implementation. Third, one could generate or facilitate the other. From what I hear, the implementations of AI and digital twin symbiosis are tracking in that order.
Manufacturing is not just building stuff, it’s also moving parts and finished goods, gauging demand, predicting supply pricing for optimum purchasing, and so forth. Every single enterprises that currently uses both digital twins and AI in their real-world processes uses some AI for this non-building-stuff activity, and their comments tell me they expect to use more. Today, just short of a majority of manufacturing digital twin users use digital twins in some aspects of transportation of material and finished goods, warehousing, etc.
This first-level symbiosis seems to be the primary driver of the second level, the process cooperation level. For example, scheduling of things involves not only the processes of moving them around, but also the timing of a completion of an order, based in part on movement, part on storage, and part on actual manufacturing. Thus, there is a value to having feedback across the boundaries of all these processes, and a means of coordinating the whole as the sum of the process parts. That role integrates things done via digital twinning with things done via AI.
A complex real-world process is almost surely not a single digital twin, but a hierarchical system of twins. As process cooperation between digital twins and AI mature, it seems to some enterprises that AI, in agent form, could become an element in the hierarchy. In other worlds, AI could create the same sort of model for some process components as digital twin technology would for others, and AI might also create a “layer aggregator” element that combined digital twins at a lower level. Right now, only a small number of highly thoughtful process planners think in these terms, but their comments are compelling. They believe that both AI and digital twin models are agents, and that the future of automation in any form is the sewing of both types of tech agents and human agents into a final, complete, business process. I think this is an exceptionally valuable insight, and also one that justifies our final level of symbiosis.
If automation is really a fundamental management of agents, then you need to agent-ize as much of any process as you can. NVIDIA Cosmos and a growing number of AI tools facilitate the creation of digital twins, which then builds agents that can be incorporated into higher-level process models that presumably could also be facilitated via AI.
The evolution of digital twin and AI symbiosis is interesting because it seems to me to indicate that it’s integrating what are often two very different “IT” activities, the edge-real-time-embedded-control stuff and the traditional information processing and analytics stuff. The enterprises who are the most thoughtful about the second-stage evolution of the symbiosis are showing signs of better integration of these activities, and that may mean either that this sort of symbiosis drives IT integration to its highest level, or demands that it get there in order to develop the symbiosis. Chicken first, or egg?
I have to wonder whether, whatever is first, we’re not seeing a need for a shift in the way we do IT. To the extent that real-time process automation has been done, it’s typically handled independent of or only somewhat integrated with traditional IT. Might we now be seeing a need to think of an entire business as a real-time, automated, system? We now have a university generative-agent-AI twin representing a thousand humans. This sort of thing could help us model entire workforces, eventually.
Does this essentially create new AI-kills-off-mankind risk? “Distributed Hal?” I’m sure somebody will suggest that eventually, but in the meantime we can always watch other humans with suspicion. That’s where our biggest threat has been all along.
