Many organisations are struggling with adopting generative artificial intelligence (GenAI), but one business already well on that journey is Vodacom.
Originally launched as an SMS channel, the chatbot TOBi was one of Vodacom’s first public-facing examples of the use of machine learning to look for certain keywords and then help customers get answers to basic questions.
Today, SuperTOBi, as it’s now called, uses GenAI to get a better sense of the customers’ demands.
“We've been developing the technology over time, but with the advent of GenAI, it got us a better understanding of customers’ intents,” says Nkululeko Thangelane, group executive head: big data, AI and ML at Vodacom.
He says while Vodacom had made a shift from keyword search to include more intent detection and semantic search, once GenAI capabilities became available, Vodacom was able to add a new framework on top of the existing stack, making the service smarter and more accurate in its responses.
And SuperTOBi isn't only a customer-facing service; thanks to the upgrade over the years, it's now integrated in different parts of the organisation and is internally facing to help agents and employees find answers.
Further GenAI adoption is found in the call centre.
“We've developed a lot of AI use cases involving transcripts to help assess if we're dealing with calls efficiently, and checking if the agent covered all aspects that were required in a certain type of call, ie, meeting compliance requirements.
“Before the introduction of this service, all of that had to be taken from a sample of calls, and a team of people would have to listen to them. Now we're doing it throughout the call centre and starting to understand what's being said, gauging customer intent.”
The automated summarisation of incoming calls is also being used to help improve customer experience by automating note-taking and adding it to the customer’s records on the CRM. This means when a customer is passed between agents, the need for the customer to re-explain their issue is significantly reduced.
“Once the call goes through, we summarise it, categorise it, determine if it was resolved or not, and highlight any action points for our team, so that the next time a customer calls in, the agent will be able to see what happened before.”
Previously the creation of such notes on the CRM needed to be made by call centre agents and this wasn’t always effective, he adds.
“That's a use case built in South Africa, that we will roll out into Africa, and we’ve shared it back to the Vodafone markets,” says Thangelane. Vodacom is the African group of operating companies allied to the wider Vodafone group.
Text summarisation services from large language models are also being used to sift through free text responses on various customer surveys used to track Vodacom’s net promoter score, adds Thangelane.
“Now, we can use LLMs to help track emerging themes that come through.”
GenAI has also been applied to improve sales propositions and move beyond keyword search across Vodacom’s digital channels, specifically its website and super-app Vodapay.
“We've taken the sales pipeline, understood our catalogues, so when we make recommendations we can make them through free text and use an LLM to understand the benefits of a product and which types of customers would get value from it.”
Thangelane says Vodacom is committed to finding GenAI use cases. With larger complements of data scientists in South Africa, Egypt and Kenya, as well as in some of the other operating companies elsewhere in Africa, he says there are around 90 data scientists across the Vodacom Group. “We've committed to 30%-40% of that resource looking at how we implement GenAI and the new tools we have.”
And what of GenAI’s impact on BI?
“I think BI is getting supercharged by GenAI,” he says.
“The models are now able to understand data. Different parts of the business have good questions about their data, but that's obviously sitting in an SQL database and needs an analyst to extract that and create a dashboard, but there’s often back and forth to get it to meet the business’ exact needs. If LLMs have access to the database, you can take free text questions and convert them into insights that can be presented, along with reasons and rationale. We're finding the models are working much better when it comes to the data.
“The conversation around BI has always been about looking at the past, and data science is about looking at the future and predicting. But with GenAI, the gap will be reduced. With GenAI, BI will get democratised a lot more than before.”
Share