Andrew Lloyd Weber has a great line that describes tech these days. In the form I’ve heard it performed, it goes “You talk of truth? Is truth unchanging law? We all have truths—are mine the same as yours?” Leaving the question of whether truth is unchanging to philosophers, consider the next point, which is the inherent subjectivism of truth. Fact is, we all see things differently. Some of that bias is emotional, some is based on the fact that we all have different knowledge sets, and some is intentional. In most new areas of tech, we see a little of all of this.
NVIDIA published a survey last week on the state of AI in telecom. Is it “truth?” I have fairly regular contact with 88 telecoms, represented by comments from 194 professionals. I’d like to compare what they’ve told me to the results NVIDIA presents. Where they are sharply different, I’d like to look at why that might be the case, and then to summarize what I think the actual state of AI in telecom is.
Of my 194 commenters, there are nine who I’d classify as persistent AI skeptics; the whole thing is a fraud and there’s little or no compelling value. Outside that nine, every single professional believes that AI could reduce opex for telcos, but only 123 thought it could increase their revenue. A total of 149 said their operator was using AI somewhere, and this represented 80 of the 88 of the telcos I’ve heard from. However, of that 149 only 37 said their use was “extensive”, and this represented 17 of the 88 telcos. Most (134 of 149) said that they used AI somewhere in network operations, and of the 37 with extensive AI use, 31 said that use was in netops.
The telco people I get comments from tend to be from the operations group (111), the office of the CTO (38), the CIO (18), marketing and market planning (12), and CEO/CFO staff (15). I don’t know how the NVIDIA survey divided organizationally or what their telco representation was, but I suspect my sample was more biased toward operations, which might account for some of the differences in view.
Another thing I can’t address with the NVIDIA survey is the kind of AI they’re talking about. My sampling excludes those who use AI for what I’ll call “personal support” missions, like document aids, emails, and office co-piloting. Did NVIDIA include these? Don’t know.
No telco type who offered me comments suggested they believed that AI could have a significant impact on their revenues. I’ve had pretty deep conversations with 15 telcos about revenue futures, and in those discussions it was clear that you have to start revenue discussions with the question of services—what are you selling? AI might play a role in fulfillment, but it can only drive service revenues if it’s what you’re selling, and telcos see AI services as a special case of cloud or edge computing services. For the former, they don’t believe they have a real shot at the market given the long dominance of the Big Three and the clear link between cloud service success and IT knowledge on the part of the seller, for any aspiring entrants. For edge computing, they don’t feel they are in any position to drive edge adoption, again because of a lack of IT knowledge and influence.
The issue with telco AI and revenue, in my view and based on what telco people tell me, isn’t much different from the issues other verticals have with AI. Across all companies, the dominant view is that AI can lower costs much more easily than it can raise revenues, and in fact most people who chat me up on AI don’t have any notion of how it could help raise their company’s revenues at all. Those that do could offer a possible suggestion to the telco world.
According to most enterprises, the revenue value of AI would most likely come from product/market planning missions. Among enterprises who offered business planning mission guidance for AI, well over three-quarters thought that AI aid in product/market planning could have a “significant” impact on their bottom line. Of the 15 telecoms with whom I’ve had fairly detailed revenue discussions, 6 believe that AI could help in service planning, 5 dismiss the notion, and the rest are undecided. It’s probably what seems to link members of these three groups that matters the most.
The six believers are all telcos who have been exploring non-traditional, meaning non-connection or OTT-like, service models. Thus, they are highly flexible in their service planning already, and are eager to get some new ideas on what might be a profitable extension to their current repertoire. While none were prepared to say they were certain AI could provide it, four of the six were exploring the notion already and the remainder expected to start this year. All of these telcos were at least willing to consider new opportunities in the consumer space.
The five skeptics say that they do not believe their company has any new revenue opportunities beyond the traditional service set, save perhaps for some tweaking of how services are billed or delivered. They don’t think AI can contribute to any tweaking; it’s a matter of “being in touch with your customer”. The operators in this group were more likely to cite regulatory issues, too, and they were focused on business services as the only potential target space. You’d think this group would be looking to AI as a cost management approach, but surprisingly they seem to be skeptics on AI overall, and thus are reluctant to admit any cost-savings mission would likely help them. Only one of this group said they were exploring AI for cost management.
The middle-of-the-road or undecided group’s common thread? Confusion? No, but perhaps resignation. They admit to having largely abandoned any new-revenue strategy beyond the notion of lowering price to gain market share. To ensure this doesn’t lower revenue, profit, or both, their interest in AI is totally focused on cost management. In fact, this group had the greatest interest in AI; three said they were already using it in an exploratory way, and the remaining one said they’d have a pilot in place in the first half of 2025.
Overall, the story I get seems almost the opposite of what NVIDIA reported. No telco told me AI was helping with revenue in any significant way, almost all said that they used AI in netops, and most of the planned use was there as well. I can’t say why these massive differences exist, but I can speculate that it’s a combination of the traditional survey bias problems. People like to look leading-edge in surveys, and almost all surveys will introduce a subtle bias in the way questions are asked and the way answers are presented and evaluated. There may also be a bit of “playing to the Street” going on; public companies are rewarded for any AI plan and dissed if they don’t have one.
I’m confident in my result here; I’ve had enough comments to be comfortable with the views and their interpretation. Overall, telcos like enterprises are struggling with AI, but AI doesn’t change the nature of a business or the nature of the businesspeople. It’s not too different from the early days of computing, when businesses knew that there was going to be a major and even traumatic change, and weren’t sure exactly how to manage it. We got through it then, and we’ll get through it now.
