Making of: AI-Driven Offer Lifecycle Management

Since the emergence of the Web and mobile browsing (who remembers WAP?), the communications, technology and media sectors have evolved jointly. New waves of innovation have come thanks to the Cloud, mobile computing, and now artificial intelligence (AI). One of the most exciting and dynamic developments, AI is a general-purpose technology, still in its infancy, but poised to alter the economic fabric across sectors, touching people’s lives in profound and unimaginable ways!

Having previously won several Catalyst awards, including Zero-touch Digital Marketplaces, Business Operating Systems and Sponsored Network Slicing, we were keen to capitalize on the opportunity to co-create an innovative solution with fresh experiences using AI, while leveraging our ODA-ready Catalog within an ODA Canvas. Before showcasing our newest Catalyst project, on AI-driven Offer Lifecycle Management for digital service providers, let me share a behind-the-scenes glimpse of the team’s efforts to date.

Creating and managing agile, personalised offers at scale

The idea for our Catalyst project arose through discussions with five leading service providers, specifically to help realise the vision of dynamically serving each end-customer – both new and existing – in real-time with personalised offers using Generative AI (Gen AI). “Not only do I know you, but I hear you, and understand your current context, expectations, and needs. Based on this, I’ve created a personalised offer that I think you’ll love”. These are the words that got us started. To help service providers manage the fast-growing complexity from this mass customisation, we also identified the need to rationalise the commercial catalog structure, ensuring efficient lifecycle management.

Developing compelling, ethical AI solutions

To bring to life our innovative use cases, we’ve combined several technologies, including a Large Language Model-based (LLM) Gen AI model running on an AWS Cloud ODA Canvas, integrated with our ODA-ready enterprise catalog and omnichannel order capture and quoting platform using the TMF 921 Intent Management API, a text and speech enabled Chatbot, and a digital customer engagement platform. Using customer and offer training datasets, a knowledge base to provide context, combined with prompt engineering, we’ve been testing and tuning the Gen AI model’s behaviour to deliver the expected results.

The model was taught to compartmentalise user interactions, ensuring privacy. “Our solution uses ‘selective amnesia’– it doesn’t have memory of other interactions. When you start a conversation, and don’t provide context of previous interactions, the LLM essentially starts from zero” explains Benjamin, our technical project lead. “In real world cases, a user must consent that their interaction data can be used to create personalized offers. For a prospect, the interactions we’ve defined don’t solicit much personal information. With minimal inputs the model can match you to an existing profile – it doesn’t need much information to do so. For customers who have a contract, this is no longer an issue.”

To provide transparency, we plan to add the ability to ask the Gen AI assistant to explain how it arrived at a given answer. Mitigating bias is another goal we have, to ensure ethically created offers. Using a Customer Data Platform (CDP) in practise should help address this point, assuming the data is broad, reflecting a diverse customer base across a service provider’s market. As the Catalyst proof of concept matures towards a commercially available platform, respecting user privacy, and transparent, fair, un-biased results, are all top-of-mind objectives.

Engagement x Collaboration = Valuable Business Outcomes

One of the rewarding outcomes of this project has been the high degree of enthusiasm and engagement from our collective team members, both Champions and technology partners alike. Everyone has been keen to provide ideas, insights, and constructive and supportive feedback throughout the process. For example, the initial ideas for our use cases were proposed by Patricia, Head of Channels and Customer Engagement at Telefónica, and Florian, Product Catalogue Manager at O2 Telefónica. These ideas have evolved based on everyone’s inputs. Many team members have also been active beyond their expected roles. For example, David, Principal Software Architect at Viasat, has gone beyond simply providing direction and feedback during development. In several cases, he’s rolled up his sleeves, lending his hands-on technical expertise to help our concept progress. “Everyone is working really well together, towards our common, shared goal. We recently advanced on three different components and now have a working solution, which we’re fine tuning.” proudly proclaims Visnja, our Catalyst project manager.

On behalf of the whole team, we look forward to sharing our AI-driven Offer Lifecycle Management Catalyst in Copenhagen!

Sponsored by: