Ray Le Maistre, TelecomTV (00:11):
Artificial intelligence is turning tech and much of the broader world on its head. But where are we exactly in terms of the relationship between AI and the development of next generation telco platforms? Well, to get some of the answers, I'm talking today with Danielle Rios, who is the acting CEO at Totogi. Dr, great to see you as ever. So look, everyone's talking about using AI to build software, but how are recent advancements in AI models and coding tools changing the way the developers actually write debug and shift software?

Danielle Rios, Totogi (00:47):
Well, Ray, I don't know how much you keep up with all these advancements, but I certainly do as I run a software company and the progress in AI coding is absolutely astounding. So we've recently had two releases of the two major models, Anthropic's, Claude and Open AI's Chat GPT. And every so often these guys report out on their SWE bench verified test, which is a test against human coding capability. And we're now at the point where Claude Sonnet four five is in the 82% range on the slide that I'm about to show you here. And on chat GPT, it's at about 75%. So these two models are neck and neck against these coding tests. And when you benchmark it against what humans can do, humans are right in this range as well. They're kind of in that 77 to 80% range. And so these models are approaching human capability.

(01:42):
You kind of couple that with the other tools that have been coming along. You've got Cursor and Replit, you have the hyperscalers with their end-to-end coding tools. And now really what developers can do is give these systems a document written in plain English describing what you want to create and popping out on the backend is a tested enterprise application deployed to the cloud. It is insane, obviously to Totogi is doing this, but the big tech companies are really leaning into this super hard. So Microsoft is reporting about 30% of their code is written by ai. Now Google at 25%. These are figures that are a couple quarters old. So I imagine it's only increasing as these models get better and better. And so at to Totogi, our developers are at the point where they can generate 10,000 to 20,000 lines of code. Previously before ai, that number was probably about 2000 to 4,000 lines of code a day. And so when you're generating 20,000 lines of code a day across your organization, it kind of just really opens up what you can create. And so I would liken the shift in coding capability as big as the shift from punch cards to object oriented programming. We're doing object oriented programming now to ai. And so I think the days of hand coding are almost over.

Ray Le Maistre, TelecomTV (03:09):
Well, I mean that's a pretty astonishing development, but I guess it's still quite early days. And it appears that there's a belief that AI coding or vibe coding as some would call it, can't actually handle telco grade applications because of the complexity and mission critical nature of those applications. Do you think AI is ready to build telco applications?

Danielle Rios, Totogi (03:36):
I do. And so I was speaking with an Asian Telco CIO a few weeks ago and she said, don't talk to me about vibe coding. It's not possible in telco. And I 100% disagree. I just wrote a blog called Context is Everything. And when you add Telco context and couple IT with these models that can now produce this kind of level of code with these tools that can really help developers accelerate their productivity, I think it's absolutely possible. What's so great about Telco is we have a ton of context lying around. We have 5G standards, three GP at Sea TM forum camera. We have tons of technical documentation, and this is perfect fuel for these AI coding models to generate enterprise level code. And so what's so great about AI is they don't get tired, they don't miss details, they don't make mistakes. And so this AI generated code can be and probably will be in the future, higher grade, higher quality than human generated code. And so this is so much more than chatbots and agents that we're talking about. We're talking about enabling telcos to create their own enterprise applications. And so I think the way to think about it isn't just ai. AI isn't just ready for telco. Telco is the perfect use case for coding because we run on specs standards and rules.

Ray Le Maistre, TelecomTV (05:05):
Okay. That's an interesting perspective there, but something you've been looking at for a while I know. And with BSS magic to Totogi taking advantage of recent leaps in AI coding capabilities to build A BSS directly from your ontology, which seems like a new approach to building telecom software. Now, you started this in late 2023. How far along are you on that journey?

Danielle Rios, Totogi (05:31):
Yeah, so I talked at MWC 24 with this vision of letting telcos create their own BSS with AI code generation. And obviously if you remember that chart, the coding, the LLMs weren't really ready to fully generate an application, but I knew it was just a matter of time before that capability was going to be there. And so we're two years in every six months, those models are getting better and better on the coding benchmarks. And so we're actually applying us to our customers as we test the capabilities of the model. So we recently did a billing migration for a tier one telco in North America. The estimate from the project managers doing it the old way hand coding would be like an eight month migration. We did it in two weeks. We recently acquired Cloud Sense who had been trying for years to fulfill the manifesto for the TM forum open APIs. They needed to do about 13 of them. They thought it would take about two months per API. We did it in four weeks instead of over two years. And so when we show this to CSPs and operators, there is an aha moment of the power that BSS magic and AI coding can do for them. And so I think the vision is clear. It's getting clearer, right? Every telco will be able to create their own BSS with ai.

Ray Le Maistre, TelecomTV (06:59):
Okay. I mean, this is a really interesting and critical turning point for telecom software it seems. So can you just walk us through how BSS magic turns your telco ontology into executable software?

Danielle Rios, Totogi (07:15):
Yeah, I think this is the question everyone's asking. Everyone's poking holes when we go show people, they're like, well, how does it work and how are you doing it? So there's an element of secret sauce, but the way we think about it is a three layer approach. And so at the bottom layer, here's what we call a zero copy data layer, right? This is a fundamental key approach where we're not requiring you to swap to togi BSS or swap to another system or move your data to a data lake and normalize it. We literally keep your systems as is including Amdocs and connect to it. And now the second layer is our ontological layer, right? This is the telco, DNA, the universal translator that once you're connected to all these systems, now understands and can talk telco, right? It's getting all the systems to speak the same language.

(08:04):
This is the vision of TM Forum. It's just approaching it in a slightly different way. And so the last layer is that AI coding layer, right? It takes the ontology, it takes your data, and now you couple it with any horizontal coding application, cursor, rept, cloud code, N eight N, any of these tools that a lot of the telcos are starting to use. And it now brings that context that you need to really produce high quality results. So what used to take years now happens in weeks. And so BSS magic is telcos universal translator. It's unifying your messy legacy, whatever you have installed, and AI can code it into a system that telcos can run themselves.

Ray Le Maistre, TelecomTV (08:50):
Well, I'm sure there'll be an equal number of people concerned and worried about this as there are people excited about it, but it's the excited people that we really want to think about and the excited people that we know that are going to be in disseldorf for the AI Native Telco forum. And you are going to be on a panel there panel titled the AI Native Telco from concept to reality. But I understand that you've got something bold planned for the event as well. What are you actually going to be showing off on the show floor?

Danielle Rios, Totogi (09:23):
Well, me, Ray, I always like to take it to the next level. And I think a lot of the CSPs that we encounter are really a little bit tired of vendors who talk a big game and then can't really produce results. And so I think a big part of AI evaluation going on is proving it. And so we want to prove it. We want to prove that AI coding works. And so I'm bringing one of my 20,000 lines, a code a day for deployed engineers out to Dusseldorf, and he's going to build five BSS modules from scratch live during the event. So this isn't slideware, this isn't smoke and mirrors, it's real AI coding. It's real telco software. And so if you're here at the event, I encourage you to stop in and check it out and watch him work and see how he can swarm code.

(10:16):
We're actually past vibe coding. We're swarm coding now where we have multiple streams of coding going on, simultaneously managed by one person. If you're not at the event and you're curious about what we're doing, please reach out to me and let's set up a meeting so we can show you what we're talking about. And so I always like to compare myself to Elon Musk. He's building his rocket to Mars. We're coding a brand new BSS from scratch to prove AI coding works for Telco. And I don't know of a better way to kind of just lay it all out there and throw down the gauntlet and really challenge vendors on. We got it. It works. Come see it, and we're going to show you,

Ray Le Maistre, TelecomTV (10:56):
Well, it sounds like show floor demos are entering a new era there. So really exciting to hear what's going on at the event. Good luck over the two days and look forward to chatting with you again soon to find out what Totogi is doing next.

Danielle Rios, Totogi (11:13):
Always great to talk to you, Ray. Thank you.