The TMForum, or TMF, is probably the body that embodies telecom conservatism more than any other body. Almost all the major telcos are members, and the attendees represent the CIO organization and the venerable OSS/BSS software that is the “core application” of telecom. BT has always been an active and influential member of the TMF, and so it’s not a surprise that a TMF interview with a BT executive is getting play on the TMF “Inform” site. What is that interview telling us about telco thinking? It focuses on what I think is a pivotal question, which is whether NaaS can present connectivity in a more profitable way. If it can’t, then telco tendency to hunker down on connection services is a big mistake.
NaaS has been a kind of hope beacon for telcos for a decade. In brief, the idea is that instead of selling fixed and systemic “service” like MPLS VPNs, telcos would sell episodic, user-and-application-specific services on demand. In practice, that hasn’t been a success, because enterprises have told me from the first that their only interest in any change to connection services would be based on a lowering of their costs. Obviously, something enterprises spend less on isn’t going to help telco profits.
BT’s Global Fabric, its NaaS platform, has tried to change up this game by focusing specifically on cloud connectivity, but in an underlying sense the focus is still on connectivity and the quality of the network. The BT executive, Colin Bannon, says “We have been astonishingly bad at marketing the importance of network, whereas our colleagues in the software world, in the cloud world – the hyperscalers – have done a remarkable job of defining reality through their lenses.”
That, in my own view, is simplistic. The cloud is a place to host applications, and applications are what empowers users and builds business cases. Bannon adds that “This silence of the telcos needs to end. A good quality, deterministic network is a prerequisite for cloud to work. And we’ve forgotten that.” OK, yes, you need connectivity for the cloud or any application to work, but connectivity is already working with the cloud, and all attempts to sell the notion that paying more for it could be justified have failed to capture enterprise support. BT seems to realize that, and now proposes a NaaS transformation built around AI.
The initial focus of the AI is back to connection properties, with things like path computation and orchestration, and possibly on to a kind of intent-based network service that aligns itself with usage automatically. This might be something that cloud and AI users would see as a differentiator, but I’m not seeing any indications that enterprises would pay for it, because they don’t currently see any way to monetize these benefits.
You can say the same thing about the interview assertion that AI drives increased network traffic. That benefit’s a bit problematic for two reasons. First, it’s not clear how valuable hosted AI services are; enterprises say that the best applications for AI have to be self-hosted for sovereignty reasons, so the “internal” model running wouldn’t be over any form of WAN. Second, few business services are usage priced, and enterprises have tended to shun usage pricing because they see it as creating a major risk of cost overruns.
The next argument in favor of AI NaaS is that the service provider edge is the logical place to run an AI model, and that this means that edge connectivity has to accommodate AI’s requirements. Both these statements are suspect IMHO.
If we assume that AI is being used to control real-time processes, then it is potentially an edge application. However, it’s a major leap to say that the “edge” here is the service provider edge. Today, we run process control systems at the edge, but it’s the premises edge where they’re hosted. No enterprise has told me they run anything at the service provider edge, or are even aware of any such capability. Why would an AI process be different.
If we did run AI at the service provider edge, would this then justify a NaaS service that responds to application intent. All such applications today are based on some static-parameter service. In a latency-sensitive application, would the time needed to assess intent and respond with connectivity even be practical. I don’t see how.
I’m sorry to say that I don’t see anything compelling here, anything that doesn’t boil down to something that’s essentially an on-demand connectivity service that’s valuable to users only if it’s cheaper overall, and to telcos only if it’s more expensive. At best, it’s simply a new way of saying something enterprises have already heard and rejected. At worst it’s another exercise in telco self-delusion, at a time when that sort of thing could be very dangerous.
The base argument the interview makes is that telcos need to sing prettier. Yes, but only if they have a pretty song to sing. If the notion of better articulation leads to the notion of having something better to articulate, then all this might be a long way around to an important point. If it’s only about the voice and not the song, then this is a signal of deep trouble.
The TMF has done good work, and has been very successful, in their basic OSS/BSS space. Is the interview, and other recent activity, demonstrated that the TMF is looking to help telcos improve their business position by identifying ways they could cut costs or raise revenues? If so, does that create a risk that the body’s membership, focused as it is on OSS/BSS, is a community that’s not fully open to new ideas and technologies? After all, it’s the office of the CTO that, in most telcos, is responsible for that sort of thing.
The TMF membership isn’t the market. Decades ago, a college student did a paper that argued that drug-induced hallucinations could meet the psychological definition of reality defined by “special consensus” meaning agreement among people with special insight or qualification. The problem is that any group can create its own special consensus, and usually does. The TMF is a conservative OSS/BSS group, and thus a creator of special consensus, but does this mean it’s defining reality?
Maybe it does, by default. There is no equivalent body representing service marketing people, and the standards groups like the 3GPP that may sort-of-represent operations or the CTO people have their own reasons for inertial behavior. I think that before the telcos can talk/sing better, they’ll need to come up with a service-evolution body who can write a market-relevant song.
