How telecom operators can unlock the full potential of GenAI

Generative artificial intelligence (GenAI) has the potential to revolutionise many industries and telecoms is no exception. At the heart of GenAI are foundational models, trained on massive amounts of data and dialogues. Examples include Large Language Models (LLMs), such as Open AI’s GPT-4 and Google’s PaLM 2 for text-based communications, but there are many models in the market now that specialize in other forms of content. These GenAI models are powerful as they can conduct dynamic human-like interactions, as well as craft original content by identifying word patterns, relationships, and the context of a user’s prompt. While these advancements are transformative, the true power of GenAI in telecom requires access to specialised data and a deep knowledge of telecom business processes – and this is no easy task. By adopting a new approach, that augments GenAI models with the right data and context, communications service providers (CSPs) gain the ability to significantly boost business productivity and elevate customer satisfaction.

Harnessing GenAI and BSS/OSS to transform the telecom industry

By leveraging up to date BSS/OSS telco data and applying telco domain knowledge, CSPs can create meaningful use cases that will add significant value across business domains. For example, customer care can incorporate LLM-based digital assistants to solve complex problems, as well as enable agents to resolve issues faster. From a marketing perspective, GenAI can quickly generate text and images for the creation of personalised promotions and campaigns. Sales will benefit from GenAI by using it to provide advice on customer requests, enabling them to close the deal faster. On the operations side, GenAI can assist field technicians by providing instant access to knowledge about networks, topology, and planning data, enabling them to fix issues or install new systems faster. It can also create network and service designs in seconds.

Using valuable BSS/OSS data, such as billing history, usage data, inventory data in combination with GenAI’s real-time data analysis and contextual understanding, CSPs can personalise interactions when discussing customer issues or responding to requests. By taking on the role of personal assistants, human agents will also benefit. In this capacity, GenAI will pull data from knowledge bases and through its understanding of customer sentiment, agents will be able to quickly provide more accurate responses and respond to customer queries in multiple languages.

However, as with all emerging technologies, operators will encounter challenges during their GenAI journey. Not as straightforward as merely training or fine-tuning GenAI models on telco data, CSPs will need to acquire new approaches and ways of thinking.

Overcoming GenAI challenges

The data needed for operators to realise GenAI’s full potential is highly sensitive. Which means training public GenAI models with open access to this proprietary data (including personally identifiable customer data) is not even an option.

In addition, a sizeable amount of telco data, such as usage and inventory is in a constant state of flux – some of which changes in real time. This makes the data unsuitable for fine tuning techniques that may work with more static proprietary data.

Another challenge relates to the accuracy of the response. To realise the anticipated benefits, GenAI interactions and creations need to be extremely accurate, however the models will only be as

accurate as the data and context it knows. Since advanced LLMs such as ChatGPT have no knowledge of the telco business and its processes, GenAI can be left with ambiguous input that is not tuned to the specifics of the telco industry, which may result in erroneous responses. Providing this know-how is essential to bridge the power of LLMs with BSS/OSS data. And it’s not just about LLMs. The telco business will require multiple GenAI foundational models that are tailored for specific business needs – including those focused on images, designs, or code.

Lastly, large, advanced LLMs that meet the quality of ChatGPT are extremely costly to build in-house and have significant running costs. OpenAI CEO Sam Altman was quoted in saying GPT-4 models cost over $100M to train with daily running costs in the order of $700k. These investments are prohibitive for many CSPs.

To overcome these challenges and create high quality responses, a new approach is needed to mediate between different types of GenAI models, users or systems that make use of GenAI, and telco data. By isolating proprietary data from direct access public models, new techniques can be used to enrich GenAI models with the right data, context, instructions and knowledge it needs – in a safe and secure way. Using this new approach, it’s anticipated that immediate GenAI use cases will be in customer care, with GenAI digital assistants along with agent support providing immediate call centre efficiency benefits, as well as digital operations technicians to automate many manual tasks of today.

With the potential of significant productivity gains across business domains, including sales, marketing, business operations, and network operations, it’s not surprising that GenAI is now a topic of discussion in boardrooms. Fundamental to the discussions is their ability to rapidly form teams to accelerate the adoption of GenAI across the business.

Capture the value and benefits of GenAI

By harnessing the value of GenAI and BSS/OSS data, operators will be able to:

1. Reduce costs: By improving first-contact resolution and reducing cost per contact, customer support costs will significantly decrease, while the quality of the customer experience will greatly improve. On the network and business operational side, staff will be used more efficiently.

2. Enrich provisioning and troubleshooting support: With GenAI, digital assistants can be used to provide real-time support for provisioning and maintenance. It can also provide troubleshooting assistance for premise-based network problems, which can be personalised based on intelligent customer segmentation.

3. Increase revenue: Through the rapid creation of business ideas such as offers, promotions, and discounts, telcos will be able to close deals faster, as well as quickly design and test new services, increasing revenue.

4. Improve prediction and optimisation: Using GenAI to produce synthetic data, a sparse data set for model training of predictive maintenance or the detection of unusual calling patterns indicating fraud will be greatly improved. Additionally, the generation of new data has a wider implication in the training of predictive models and improving the optimisation of systems.

5. Deliver exceptional customer experiences: The data GenAI pulls from the operator’s BSS/OSS will result in higher net promoter scores, enhanced customer satisfaction, and improved customer effort scores.

For operators to unlock the full potential of GenAI, integration of their BSS/OSS, as well as developing new approaches and ways of thinking will be key. With its extraordinary capabilities, GenAI is set to transform and revolutionise the telco industry.