Sponsored
Gil Rosen, Chief Marketing Officer at Amdocs, shares the company’s radical approach and activities around GenAI and agentic AI at DTW Ignite, from the coming telco world without customer complaints to personality engineering for agents.
After ChatGPT went mainstream, some organizations were waiting to see what happened. Now, it's more of a race mentality as we watch hyperscalers like Google, Nvidia, OpenAI and others rolling out new capabilities rapidly and see the potential for various use cases being proven.
GenAI is a high priority for everyone, but it takes time to make really good use of it within an organization. The large language models (LLMs) that have been introduced in the last two years are from external information sources, so they have not been able to penetrate and connect to telecoms’ information systems, which are a ‘closed garden’.
Our first priority was building a solid foundation to connect to all the different domains within a telco and connect them to each other, although some were never connected before. This is needed to give call center agents sufficient insight because a customer’s complaints might involve information from the network. The right architecture must be in place before a telco can use GenAI to fully assist an agent or interact with a customer.
That architecture brings big benefits. For example, working with a very large North American operator, we deployed a PoC in a call center as an assistive function and reduced call handling time by 63% and improved first-call resolution by 50%, which led to a 50% improvement in the Net Promoter Score (NPS). More and greater benefits are accruing as we move to a wider commercial deployment.
This is a bright signal for all operators on the customer-facing side, and there’s more. We did research into how consumers feel about interacting with GenAI, and about 80% of respondents are positive about it because they already interact with it in their everyday lives.
Communications service providers (CSPs) are taking longer in the network and other domains because the stakes are higher, although they have been using AI in the network for a long time. When ChatGPT hit the mainstream, nobody realised the advancements would be so significant. Now we understand the scale of capabilities will continue to improve, making implementation table stakes. We’re in a good place, and it's just going to get faster.
People are realizing that if they don’t take a proactive approach, it will impact their future. They are also realizing this is not a technology revolution. It's a holistic, end-to-end revolution that is changing how organizations are set up and their scale. In 10 years, will customer care be handled by 20,000 people who follow the sun? No. The real question is, how fast will we get to the minimal required number of people? Also, what does that mean for care organizations or CSPs’ network organizations, or marketing?
Telecoms is a people-intensive industry, so there is scope to make our industry much better. No more excuses for failed first-call resolutions or for customers waiting on the line. Theoretically, there will be no issues because they will be solved before complaints come in; customer care will be super personalized, and so will products.
Our CSP customers will benefit immensely too: It’s not like the cloud revolution, which allowed us to become more agile and cost-effective, but we are still selling the same products and have many of the same issues. With GenAI, we’re climbing the ladder of customer satisfaction to reach new heights – exciting for an industry with one of the lowest NPS scores. In three to five years, people should be super happy with their CSP engagement.
Amdocs demonstrated agentic AI to customers on our booth at MWC in 2024, which created a lot of interest. We used a couple of random profiles for the agents and from the interactions, we developed a new field we call personality engineering because we saw that we need to teach agents how to behave in keeping with a brand’s culture, to reflect its values and use its language.
Human agents adapt their behavior to the person they are speaking to, with agentic AI, we don’t want the same agent talking to a 16-year-old girl as to an 85-year-old man as those people don’t use the same language or approach. Further, from research we commissioned, we learned that preferences vary from place to place: In most of Asia, people prefer to engage with an older person, but this is not necessarily Europeans’ preference.
These issues are not yet being widely discussed, but they are very important. Agentic AI is not a tech upgrade, but a whole new channel for brands, and especially important for CSPs, which rely so heavily on humans.
At DTW Ignite, Amdocs is participating in three Catalysts, which are unique opportunities to partner with CSPs and have real conversations, addressing real problems.
In the first Catalyst, BIND (for Bridging intelligence, networks, and digital twin) Amdocs is working with Vodafone and Telstra, creating a digital twin for the network in real time, connecting the assurance inventory in the knowledge base. The scenario is that an operator wants to launch a new service and ensure all the components are available to provide the service. Also, if there is a failure, how exactly to fix it.
AI-powered billing QA and anomaly detection is about avoiding bill shocks. They happen because, traditionally, telcos collect information throughout the month and produce a bill at the end. Things go wrong for many reasons: Operators themselves lose about $40 billion annually from inaccurate billing or revenue leakage. Here, GenAI is applied proactively to assure quality and deal with issues as they arise. We can even predict customers’ complaints and the likelihood of them churning.
Growing B2B with autonomous agents enables flexible, fast quoting for medium-sized enterprises, which is difficult because it involves a lot of handovers, information and logistics. Currently, 50% of requests for tender have errors. We want to transform the B2B sales process using autonomous agents.
Here an agent creates the package using autonomous agents to pull information together, then it manages the proposal’s activation and ensures fulfillment, which is super important. One of the worst statistics in our industry is the lag between order and completing delivery, which often takes six to 12 months or longer. This also means a long period between concept and cash.