About Network APIs and Agentic AI (with Andrew Collinson and Dean Bubley)

There is a ton of talk about how AI is disrupting everything and will reinvent industries, especially telco. 

Is that truly the case?

At the forefront of this notion going into MWC and Ericsson’s pre-MWC Analyst Briefing are the hot topics of agentic AI and network APIs, in particular, global network APIs. 

Leonard Lee is joined by Dean Bubley of Disruptive Analysis and Andrew Collinson of Connective Insights to unpack the potential and the unthinkable possibilities of a fusion of network APIs and agentic AI at a heavy metal gastro pub, World's End, in Camden Town, London.

Hit Leonard, Andrew, and Dean up on LinkedIn and take part in their industry and tech insights.

Check out Dean and his research at Disruptive Analysis at www.deanbubley.com.
Check out Andrew and his research at Connective Insights at www.connectiveinsight.com.

Please subscribe to our podcast which will be featured on the neXt Curve YouTube Channel. Check out the audio version on BuzzSprout or find us on your favorite Podcast platform.

Also, subscribe to the neXt Curve research portal at www.next-curve.com and our Substack (https://substack.com/@nextcurve) for the tech and industry insights that matter.

NOTE: The transcript is AI-generated and will contain errors.

DISCLAIMER: This podcast is for informational purposes only.

Transcript

Copy

Andrew Collinson 0:06

Next curve.

Leonard Lee 0:10

Hey everyone, this is Leonard Lee, executive Analyst at Ncur, and welcome to this episode of Next Curve's Rethink Podcast. And this is a really special episode because I have disruptive Dean here, and most of you'll know him as the most. The analyst in the entire,

Dean Bubley 0:30

one of the most destructive,

Leonard Lee 0:32

no. Come on man.

Dean Bubley 0:32

Have you met John Strat?

Leonard Lee 0:34

Uh, no. Okay. You're worse than him.

Andrew Collinson 0:37

he's destructive.

Dean Bubley 0:38

Oh yeah. Yeah. He's,

Leonard Lee 0:39

he's worse than you. I mean,

Dean Bubley 0:41

well, not worse. He's not quite worse. Different,

Leonard Lee 0:43

you mean he's better than you?

Dean Bubley 0:45

So he goes in with a Viking patch and I much prefer as Smile and St. Jagger.

Leonard Lee 0:51

Okay.

Andrew Collinson 0:52

Things like precision disruption.

Leonard Lee 0:54

Ah. Precision disruption as opposed to wanting,

Andrew Collinson 0:59

as opposed to nuclear color.

Leonard Lee 1:00

Nuclear. Mutual annihilation.

Dean Bubley 1:02

Okay.

Leonard Lee 1:02

Well, you know, throwing rocks is not entirely a. Thing. And of course we have Andrew Collinson here, and these two gentlemen make up the

Andrew Collinson 1:12

unthinkables

Dean Bubley 1:14

absolutely unthinkable,

Leonard Lee 1:15

the untouchable unthinkables. And of course, with me it becomes ridiculously unthinkable, right?

Dean Bubley 1:23

possible. Impossible.

Leonard Lee 1:25

Yeah. Impossible.

Andrew Collinson 1:26

You are very welcome to join us.

Dean Bubley 1:27

I'm welcome to London as well.

Andrew Collinson 1:29

Yeah.

Leonard Lee 1:29

Cheers to that. We're

Dean Bubley 1:30

here on a Sunday afternoon in a proper rock and metal paring candy.

Leonard Lee 1:36

For many of you who have been following Dean for years, you probably don't know that he is a metalhead. Yes.

Dean Bubley 1:46

Sometimes.

Leonard Lee 1:46

I mean, why are we here? Is. Because of you.

Dean Bubley 1:50

Right. 'cause of the YouTube and the guitars. So you've got your guitars down there. Yeah.

Leonard Lee 1:54

And hey, Andrew is a luthier

Andrew Collinson 1:58

dean. I am. Yes.

Leonard Lee 1:58

Yeah. Which is amazing. He's also a accomplished guitar player,

Andrew Collinson 2:03

allegedly.

Leonard Lee 2:04

Yeah. Well, who, hasn't practiced in long enough. So he's lost his callus.

Andrew Collinson 2:09

We gonna now, we're gonna give it a go tonight then.

Leonard Lee 2:11

But we're not here to talk about how, unthinkable and ridiculous we are and our, about our guitar playing in our guitar making. We're here to talk about network APIs and spectrum policy and God knows what, but it has to do something with.

Andrew Collinson 2:31

Really You want talk

Leonard Lee 2:32

about that? Yeah. Telco, right? Sunday

Dean Bubley 2:33

evening

Leonard Lee 2:34

s

Dean Bubley 2:34

here.

Leonard Lee 2:35

Yeah. I know.

Andrew Collinson 2:35

down the full pro pretends.

Leonard Lee 2:37

Yeah, I know it, it is kind of ridiculous. But then we did say this, I mean, this is partially unthinkable and Indeed is kind of ridiculous.

Dean Bubley 2:45

So go ahead. I mean, what do you wanna know?

Leonard Lee 2:47

Well, on Tuesday we're gonna be hang the folks at Ericsson and you guys, recently did some work,

Dean Bubley 2:53

Yeah, we did a workshop. We did, a workshop on, network APIs, meet Agen ai and in fact, one of our sponsors was aa, which is, the Erickson and 13 Telcos jointly owned, aggregation API aggregate.

Leonard Lee 3:09

Yeah, let, let's see it.

Dean Bubley 3:10

it?

Leonard Lee 3:10

Okay. he'll show

Dean Bubley 3:11

it to you

Leonard Lee 3:11

later.

Dean Bubley 3:12

that day was really interesting because if we were working on the question of whether Agen AI ends up being the potential customer for network APIs, as well as perhaps helping create them, whether it's end up is gonna be agents that are gonna be invoking APIs. Based around intent. So it could be around security of a transaction of sort of almost like say agents, validate this person's identity and, rate them for fraud and credit risk. or it could be more network related around. Well, I'm, I'm, I'm, I'm a famous slicing skeptic, but there is, you know, possibility of asking, for. You know, location information or network state information and so on.

Leonard Lee 3:58

Yeah.

Andrew Collinson 3:59

It also raises the possibility, perhaps of a different model of discovery. Yeah. You know, in the sense that if you just have very static, as it were APIs that you have to be directly connected to, we wondered if, a agent, API plus stuff like MCP would allow people to. More easily discover on a more ad hoc basis. Yeah. APIs, I mean, the answer to that is not immediately. Certainly, it's a lot harder to do from what we heard.

Leonard Lee 4:24

Yeah.

Andrew Collinson 4:24

Yeah. And indeed the whole thing I'd say is it's harder than it looks.

Leonard Lee 4:29

Yeah. And recently, I think. Nokia is working with Telefonica, right? Yeah. They're exploring some of these topics that you guys Yeah, yeah, yeah. Absolutely.

Andrew Collinson 4:38

There are real world examples of it. I mean, there's examples of, in China, for example Yeah. A couple of provinces where they've used de Gente with MCP and they guys, to deliver services. Yeah. Now whether they're, proof of concept or proof of failure, I'm not sure. It's hard to tell, but it's definitely been tried.

Dean Bubley 4:54

Yeah.

Andrew Collinson 4:54

Tele FICA launching stuff as well. I think

Dean Bubley 4:56

There's lot of trials going on. I heard about. Also FCP when I was at the, amplify Conference on network as a service a few months ago. And I think one of the interesting things here, and actually at the workshop we had is that although the focus of a lot of analysts and a lot of industry is on the mobile APIs, but hear a lot more in Barcelona in a couple of weeks time, there's a lot that's going on in the fixed world and in the broadband world and in the, the sort of the core of network. Support side of things, even Snap line.

Leonard Lee 5:27

Yeah.

Dean Bubley 5:27

And I think that, you know, part of the challenge around APIs is sometimes the over focus of, oh, it must just be on the, a cellular operator. And that's important part of it. But there you have to have, you have, you need network coverage. You need to. Enough for the operators in a given market, possibly through the MVNOs, which is something I'm looking at on the fixed side. There's more, should say mechanistic and deterministic use cases. So it might be a case that let's say you are an enterprise and you've got a factory in just law and a data center in Ohio and you are firing up a new AI process. You need to train it. With a data center somewhere else. So you might say, okay, I need to get a petabyte of data from here to over there, train this on, on those timeframes, then shut it down afterwards. there isn't one network that covers all those places, so we will have to have like a tail circuit on this other provider and we a. That sort of the provider provider piece as well as what we want and actually that type of thing is a nice deterministic problem. Whereas, yeah, we are in the, we are at like two brick walls from the outside here, and so if we wanted to do real time uplink of this, this audio, we would have to think, well, is there decent enough network company enough? All the rest of it. And there's a bunch of physical world non-deterministic problems.

Leonard Lee 6:56

Yeah.

Dean Bubley 6:57

Anyway, let's, cut off that.

Leonard Lee 6:58

No, that's good, good insight. So this is all the, key takes that you guys out. That's

Andrew Collinson 7:03

interesting. But I think if you step back a little bit and you just talk about APIs, the industry is still a little bit, unclear about how difficult it is for client. I mean. There's basically, APIs are not on anything, not APIs are on, on their own, are not necessarily that hugely inspiring or important.

Dean Bubley 7:22

Yeah.

Andrew Collinson 7:23

the ones that are being exposed can do useful stuff. Yeah. But one of the things we've been looking at is, well, that's just how complicated it's for a client to actually implement this. But they also, they, they've already implemented a whole bunch of other things.

Dean Bubley 7:34

Right.

Andrew Collinson 7:34

So if you look at a fraud, they've got their own measures.

Dean Bubley 7:37

Yeah.

Andrew Collinson 7:37

And, and therefore you are, you're kind of, you're not. Suddenly presenting them an entirely new solution. You are presenting them a part solution to a problem they've already Been dealing with for many, many years.

Dean Bubley 7:48

Yeah.

Andrew Collinson 7:49

And it's not always that easy for them to, yeah. Look into it.

Dean Bubley 7:52

And so a lot of this I think, needs to be wrapped up into IT solutions and products. Rather than selling an API, you sell an API enabled solution. And so if either an operator, an aggregator, or a specialist. In say, transaction management goes and says to a credit card company or retailer, we have this network, API enabled thing, which is 27% better than that whole thing you used, or it adds better granularity for this sort of customer, then that's good. But if you go away, say, Hey, we've got an API. Well, you and the other hundred people have phoned me up this

Leonard Lee 8:27

Yeah, yeah. And I, I, I think one of the problems the industry has is they think of API and ice. You know, like each specific API and then you're talking about the use case for that API. And then you get lost in sort of this siloed, narrow view of what the API does. But then also, you get disconnected from all the complicated stuff that happens underneath to actually surface the service.

Andrew Collinson 8:59

Yeah.

Leonard Lee 9:00

From the network. Right. And like you're saying, oh. APIs have been around forever. And in the context of fixed, in these other domains of telecom, because, you know, Dean likes to harp on the fact that mobile is not the only part of Telco. Right. Well, right. But you're, you're absolutely correct.

Andrew Collinson 9:19

Yeah.

Leonard Lee 9:20

Right. But then I think network APIs, the reason why there's, a lot of hoopla around it is because it's solving a problem that I think a lot of fixed networks don't have to really contend with. Which is how do we, how do we support the global regime, you know, like the standard regime of, mobile wireless, right? The fact that we have a single standard instead of whatever, five, six, whatever ridiculous number. In the past, remember the days when we were carrying around three, four cell phones, right? C one PCs, C DM A, right.

Dean Bubley 9:55

B, c dm A, whatever the Japan was using this year.

Leonard Lee 9:58

And so I think it, it has a different value proposition, which I think will serve as a, an. Let's say an additional mechanism to keep things together, right? In a, global regime that seems to be fragmenting at the moment due to geo quality.

Andrew Collinson 10:20

Yeah, I mean, you can argue there's a, there's a sort of, let's try and make, try to useful is one of the, there's one of the drivers of that.

Leonard Lee 10:27

Yeah.

Andrew Collinson 10:27

I mean, my, my feeling about it more than anything is it's now a credibility test. 'cause it's really about how they. You know, 10 operators prove that they can participate in that part of the data economy. The revenues matter, but I actually think the proof that they can do it matters more. 'cause nobody knows what the revenue's gonna be. Look at the, the models really hard to model, and do some work on that. You know, look at other people's work on it. It's. But it's more about be credible. We've talked the industry, talked about it for five, six years now that have gateway stuff. You really need to start seeing some traction, some numbers.

Dean Bubley 11:02

Well, I'm gonna pick up on then on something you just said. This is three. No, I'm actually not. I'm gonna check my God. Steer the conversation on what we are working on next,

Leonard Lee 11:12

what is happening,

Dean Bubley 11:13

geopolitics and sovereignty. And you mentioned geopolitics. And I think that that is a huge deal, not just on the network side, but on cloud, on ai, on all sorts of areas. And I think that one of the things we are gonna see over the next six to nine months is a bit of a pivot from. Some of the sort of in the weeds discussion around APIs and even some of the sort of automation and AI stuff too. Sovereignty is absolutely at the top of policy makers, regulators, and increasingly investors agendas. What does it mean? And how does it affect the rest of what we're doing in the industry? Yeah. 'cause there's so many touch points and, and we've got it looking into that.

Leonard Lee 12:00

Yeah.

Andrew Collinson 12:00

It's really complex.

Leonard Lee 12:01

the challenge is, is, you know, when you talk about sovereign X, Y, Z, Hmm. When you're looking at, you're looking at a regulatory regime that you need to transverse in order to get to a global scale that we've enjoyed with 5G, arguably, you know, 4G sort of. But I think that is. The problem to be solved. It's not an apparent problem at the moment. It's something that you're now, you're suggesting you're attaching to sovereign, especially with all this talk about sovereign, but I mean, literally the future is six G is in question that you may or you Yeah. Where you know, let's think about it from a standards perspective, from an innovation and invention perspective, everything is gonna be driven by a market that. Is able to deliver the most compelling capability. Hmm., What is the network API about? It's about exposing network capabilities and standardizing it using these kamara. API definitions, right? They're not the actual service, but at some point, the most mature networks and the technology staffs and the technology portfolios are, they're inevitably going to win out, whether it's through the 3G PP process. Or we're looking at a bifurcated or even fragmented future. I mean, working, basically going backwards, right? I mean,

Dean Bubley 13:33

I mean, you could argue that there have been niche versions of the cellular standards, even when we've had what we thought was GSM, that was GSMR for railway, for example, or the special versions of 5G that get used in defense and public.

Leonard Lee 13:49

five GAA.

Dean Bubley 13:50

There's, for rail, there's a thing called F-R-M-C-S. Although as I heard from a a, a conference Y last week, someone said there's no G in F-R-M-C-S. So whether it's 5G or 60 standard, that was sort of someone at knock out. I think, I think it might be Neil Connolly. I need to double check his name, but I was like, that's a great night

Andrew Collinson 14:11

To pick up sovereignty. One of the things we've kind of been talking about a lot. The sovereignty is being sold as an absolute, as a binary solution. So it is a sovereign solution. The reality of sovereignty is like a quality, so like, how tall are you? Or, you know, how, how strong is this beer? Because you, it's almost impossible for most countries and most recent to have a truly sovereign solution managed all in a connected world. Yeah, because it's like a paradigm. It's connecting. How can it be sore? It's how sovereign eye.

14:43

It's

Dean Bubley 14:43

connected and layered. So if you think about, say six G. Yeah. You've got certain things. For example, we're expecting six G to be AI native. Yeah. Whatever that means. I'm try, I'm writing a report what that actually means. But, but what you're finding is that that AI of regulation is itself regionalized. Yeah. So there's very different sets of policies in the us. So the EU and UK and Japan and China. So you, you are everyone's who's positioning AI as a core enabler of six G is immediately bringing up that fragmentation issue. Yeah. But then you have things like, even something sort of relatively minor. For example, cellular networks need to have ultra precise timing signals for synchronization of the. Station so they don't through each other.

Leonard Lee 15:30

Yeah.

Dean Bubley 15:30

And at the moment that timing synchronization is delivered by GPS in a lot of cases, but GPS is both US owned and increasingly becoming liable for jamming and spoofing.

Leonard Lee 15:41

Yeah.

Dean Bubley 15:41

So you've now got countries developing their own timing networks. Yeah. Timing solution. So you know, you end up with the fundamental building, whether it's. Timing, whether it's ai, whether it's semiconductors, whether it's the memory, the whole supply chain, you know, is itself non becoming software in its own way. And so inevitably six G, which is the amount of all of that, is gonna be driven by. Yeah,

Leonard Lee 16:07

I think there are absolute like that, that's an interesting point. your point about layers and the challenge probably with 60, is how to hold the ship together. Right. and then, but see that's where one of the, you know, like I know you guys had a lot of conversations with the doing. SI, that's one of the, potential essential functions that the a doing a exchange can play is to hold at least a few of the layers together, or at least provide a mechanism for things down, let's say at. The maybe a layer of three, four, and then maybe five together, and then the rest of it, it is a spaghetti mess anyways, right? But maybe trying to, at least provide that mechanism, cohesive mechanism. In the absence of that things can unravel pretty quickly because you know, I mean, if everyone's, you know, to the point you made, Andrew, if everyone's going sovereign, that means everyone kind of has a different definition of the policies associated with the stack. No, it's now. Transcending what we're used to. Yeah. With what the 3G PPP project has

Dean Bubley 17:25

so last year actually the event we are going to, on Tuesday, the Erickson event, I asked a question which caused a fair amount of Constellation. I was like, well, here's a do not, which is an aggregator of APIs of different operators. Can you aggregate APIs? From networks based on Chinese vendors. And can you aggregate and, and if you're the US developer, can you invoke those, which presumably while sending data to them Yeah. Of those networks. And I, I sort of, I've had a few answers, but I'm not a hundred percent sure that that, that anyone's really bottom that out.

Leonard Lee 18:02

Yeah.

Andrew Collinson 18:03

I mean, would you really, really think Adina would hold. Six G together. I mean, Jan, me, there are so many other questions that

Leonard Lee 18:09

it's all based on, governance and being able to foster trust within the network of participants, the operators, and then being able to, you know, dynamically and scalably administer the compliance. And so if you're talking about all these different sovereign regimes that you have to kind of orchestrate across, like let's say there's a particular workload or a request that comes through, how do you secure it? How do you, ensure compliance? How do you ensure that it's trustworthy and that consent is, is

Dean Bubley 18:50

governance,

Leonard Lee 18:51

you know? Yeah, it is. Enforced.

Dean Bubley 18:54

It's govern governance and consent. Yeah. And I think this is part of the huge, the value that, that they, the aggregators is both part of the value and part of the problem,

Leonard Lee 19:03

yeah.

Dean Bubley 19:03

Is that consent layer is

Leonard Lee 19:06

shocked. It's really hard. And then it's also that layer that. Completely thankless and nobody wants to really do because it's so complex. And then when we consider how fluid the geopolitical situation is and how fluid the, the, sovereign x, y, z conversation is, I mean, how do you intercept any of this stuff? I mean, it's like a thinkless job.

Dean Bubley 19:31

X, Y, Z sovereign alphabet in this country. Mate, no, there's none of this scene nonsense.

Leonard Lee 19:38

Remember where you are? Okay. Forgive me. That was brilliant. No, that's really interesting.

Andrew Collinson 19:49

Big precision disruption.

Leonard Lee 19:50

Yeah, I know.

Andrew Collinson 19:52

Non-nuclear

Leonard Lee 19:53

precision. He's just putting me in my place. That's all he is doing right now.

Andrew Collinson 19:58

Just ignore it.

Leonard Lee 19:59

But hey gentlemen, you know, it's really great hanging out with you. thank you so much for coming out and keeping me company.

Andrew Collinson 20:08

a great pleasure.

Leonard Lee 20:09

It's been good. Thank you. Thanks. Your round max as well. Yeah, I'm sorry it's your round of drinks next. Oh, yeah, yeah, yeah. Okay. You didn't hear that by the way. This is water yellow. Yeah, that's why it says foamy water. That why it says cabinet Brewery. Yeah. That's why it says beer

Andrew Collinson 20:27

tea. Yeah.

Leonard Lee 20:30

So, cheers, gentleman up. Thanks. And, to the audience who, bothered to listen in, number one. on behalf of Ncur, I really have to thank both of you. you know, I follow the work that you both do and I really appreciate it. Yeah. and I think that audience will appreciate what we just had in terms of an exchange and, yes, disruptive debate. That is what we need in the industry. And when we get disruptive, unthinkable and ridiculous people together, you know,

Andrew Collinson 21:05

oh boy, that's.

Leonard Lee 21:06

Magic can happen. Right? It's not like we're equally or in the same dimensions, Mr. multidimensional, you know that slide? Yeah. Yeah. I know. That

Andrew Collinson 21:15

know point is, it's important to disagree because is only through disagree that you find new things out.

Leonard Lee 21:20

Yeah. I mean, if we all agree, that would be pretty freaking boring.

21:23

short

Dean Bubley 21:24

podcast as well.

Leonard Lee 21:25

Yeah. Yeah. Yeah. And so remember to follow Dean, disrupt. 50 at, disruptive. Oh. Which

Dean Bubley 21:33

my company's called Disruptive Analysis and LinkedIn and Substack and YouTube and all the rest of it.

Leonard Lee 21:38

Yeah, yeah. I, and then Andrew Collinson.

Andrew Collinson 21:42

Yeah, just me, Andrew Collinson, connective Insight. So disruptive. I'm connected.

Leonard Lee 21:46

Yeah. Yeah. I mean, so he's the Yining, he's the yang, or is the other way around.

Andrew Collinson 21:52

He's the dynamite.

Leonard Lee 21:52

yeah.

Andrew Collinson 21:53

I'm the rebuilder.

Leonard Lee 21:54

Oh, okay. So you, you reconstruct everything after he is blown everything up.

Andrew Collinson 21:59

And it's like it goes round, you know?

Leonard Lee 22:00

Remember that? Yeah. And remember to follow Next Curve, www do next-curve.com, for the tech and industry insights that matter. And we are signing off from, what's the place called? The World's End. World's End. Here in Camden Town, London, UK. Take care everyone.

Andrew Collinson 22:23

Cheers. Cheers everyone.