CXO Forum 2024 Sponsor Series: In conversation with Mr Doron Sterlicht, Amdocs

Generative AI is a hot topic in the industry right now, but implementing it in a real world scenario comes with its own set of challenges. Doron Sterlicht, Head of R&D at Amdocs, a Platinum sponsor at our CXO Forum 2024, shares more with us about Amdocs’ GenAI solutions, how Amdocs could help telcos overcome adoption hurdles and how collaboration in the industry could help overcome some of these challenges.

Ken: Hi Doron. Can you share some of the key obstacles companies are facing when trying to deploy GenAI solutions, particularly in the telco sector?

Doron: This is a great question. Because, you know, 2023 was the kind of the era of exploration and 2024, the title I would put for 2024 is the year of Realistic Expectations, where many organisations were trying to harness this disruptive technology, but they faced many roadblocks. Among them, it’s the hallucination where the AI invents things because of lack of context, limited reasoning.

There were hopes that it can do so many things, but then a lot of organisations discover that they need to do a lot of development around the GenAI to actually produce those good results, the security issues. And even if you managed to overcome all those challenges, the performance and the cost was a big challenge because these models can be very expensive to operate.

And specifically for the telco. What we’ve seen working with many telcos around the world is challenges around understanding the domain of the telco – how to create models and applications that can truly reason and perform action that can do more than just Q&A on some index data. How do you use that data? How do you organise it?

How do you sometimes synthesize data from a small seed of data? Because you can’t just go and use the data of your customers. Another challenge is integration, that is all the time and problem. But with GenAI, it’s an art in itself. How do you massage the data? How do you feed the data in a format that is very complementary for the AI to actually use…obviously, security, privacy and regulation, we see a big need to deploy some of the use cases on-prem and to have the necessary, security layer on top.

And the last point is development knowledge and tools, where basically, you know, how do we allow democratisation of using GenAI in organisations, especially local tools, and the rapid development frameworks.

 

Ken: So given these challenges, how is Amdocs helping telecom companies and other industries overcome those barriers? Can you share some specific strategies or solutions that Amdocs is providing to ensure successful GenAI integration?

Doron: Absolutely. It’s a great question. Last year we launched amAIz – amAIz is our GenAI platform. It carries a lot of platform capabilities, such as catering for the privacy and securities. We can run both as a SaaS and on-prem. It includes all the necessary components required for our chips and trusted AIs, such as guardwell, topicality, toxicity and protection against a whole new family of security threats in the GenAI. They learn how to manage tenants, to monitor consumption and reports, and to even build any use case you may want in the organisation.

But on top of that, it also comes with ready made, telco use cases that are manifested in what we call the telco agents. These are basically a set of digital employees around billing, care, sales and network automation that, can actually perform a wide range of tasks. Ultimately, being able to handle a very complex query by the user reflecting a call from start to finish. These agents, each one of them has different telco skills.

These are encapsulated capabilities that otherwise a human would need to have. And they all work in unison to give a holistic, seamless experience to the customer experience. So the first query might come to the AI entity responsible for the billing care, because the customer is asking, hey, why my bill is so high? And then maybe, you know, a promotion got expired.

And the AI explains that to the user, but then the user wants to have okay, you know, but this is too expensive for me. Maybe I want another plan. Or maybe we need to offer him a new promotion. And then it goes to a different AI entity behind the scene responsible for sales and commerce. But from customer perspective is talking to a single entity, and that single entity, is providing all the support the customer needs.

We are also able to identify if the AI is exhausted and if a human needs to take over. And this is, a big part of what we are offering on top of the platform capabilities.

 

Ken: Are you able to share a customer use case story where Amdocs has effectively implemented GenAI? What were the key outcomes and how did it transform the customer’s operations?

Doron: Absolutely. So, actually, we recently deployed one of our billing care agents in a North American account.That agent is largely powered by experts model that we work together with Nvidia, to try and build. From a technical perspective, we managed to achieve around 80% improvement in the latency, 63% cost reduction due to the fact that we managed to reduce the token consumption dramatically and over 30% increase in accuracy. Leveraging on the enhanced reasoning and we managed to, build as part of the product. Just to give you an ideal way to get to that results, we had to analyse hundreds of thousands of transcripts and to really get to the root cause of every customer query and extract that reasoning and place it back into the application.

Some in the model, some in the wrapping application. But in the end, we managed to get to that level of improvement – so these were more like technical KPIs. From business perspective, the agent was able to answer complex billing inquiries. That again, truly mimics human behavior. And going after the root cause. And, we managed to get more than 63% reduction for the average handling time, 50% in first call resolution, and almost 49% improvement in NPS, which are tremendous results.

We’re very happy with it. But it’s just the beginning, because what we enabled on that project was just a subset of the telco skills that agent has. And now the project will evolve. We will unlock more capabilities both around the billing care as well as around other telco domains such as conversational selling and more.

 

Ken: As we continue to explore the potential of GenAI, what role do you think collaboration between tech providers, telcos and other stakeholders will play in driving innovation and overcoming future challenges?

Doron: It’s a great question. I think it’s already happening. We see, like ourself working very closely with Nvidia and Microsoft on the amAIz Platform. We see others doing something similar, and I believe partnership is really a key for success. The industry will face a lot of challenges that are just too great to be solved by a single company, and also a big need for standards; standards around security, standards around communication protocols between AI and AI, because we’re going to see more AI actually communicating with the AI, whether it is a personal assistant of a consumer that is actually speaking to AI of the CSP, or even something which is completely kind of machine-to-machine, like, autonomous cars and so on.

Also standards around how to drive all that generated content efficiently over the network, leveraging on AI on the edge. So we’re going to see a lot of examples around that. And maybe a great example is, you know, connecting to the previous point is the initiative that we’ve seen in the market of sovereign AI factories where a telco, especially large groups, are partnering with technology companies and ISVs like Amdocs to create new revenue stream, to create those factories that will empower different enterprises and governments, within the region. And really like it’s a whole new business for telcos that are very well positioned to take it back.

 

Ken: Thank you so much, Doron, for a very candid conversation.

Doron: Thank you very much. Very glad to be here.

Ken: That was Doron from Amdocs. Thank you for watching ‘be’ in conversation.