Vodafone Group CTO Scott Petty (third, from right) laid out how AI has the potential to fundamentally change the economic landscape of the internet, but developing markets will be slow to see this change, due to challenges such as low smartphone uptake and lack of fundamental infrastructure to drive the resource intensive “hyped” technology.
Speaking at a media briefing in London at Vodafone’s global headquarters, the chief technology officer called generative AI “probably the most overhyped technology for many years in our industry” despite its landscape shifting potential.
“Hopefully, we’re reaching the peak of those inflated expectations because we’re about to drop into a trough of disillusionment as companies really struggle to go from POCs [proof of concepts] from gen AI and create real scale benefits across large organisations,” said Petty.
There are a host of challenges that make tapping into gen AI “difficult to do”. Petty detailed Vodafone is adopting a “partnership model” whereby Vodafone taps the “largest hyper scaler producers” of AI, instead of building from the ground up by installing graphics processing, installing them in data centres, and building its own large language model (LLM).
Vodafone announced in January a 10-year strategic partnership with Microsoft to leverage AI and enhance its operations across its global footprint.
This is the right approach he argued as the latter would leave the operator perpetually behind as the technology is rapidly evolving. Also, tapping into public LLMs such as Open AI’s platform would expose the operator to data privacy challenges.
Vodafone developed and “fine-tuned” LLMs for generative AI applications in multiple areas to streamline business processes, boost employee productivity and upgrade its chatbot TOBi.
In total, Vodafone has built 600 models with traditional AI and machine learning for various use case stemming from financial modelling to network prediction, and they have “delivered a lot of value”.
The key to getting gen AI right is quality of data, said Petty. Vodafone has been pooling its data ocean since 2016 which now totals 24 petabytes. By doing so, LLMs can deliver accurate answers and avoid AI data hallucinations, which is when a LLM reads a pattern of data to give nonsensical or inaccurate answers.
Vodafone developed and experimented with an AI-powered FAQ search tool for contact centre agents which only produced results at an accuracy rate of only 25%, this was how Vodafone recognised its data quality was “crap” said Petty.
“We hadn’t been lifecycle managing so we were answering iPhone 15 questions with iPhone 13 data, we weren’t managing the infrastructure that goes behind that. So a key critical success factor in generative AI is your data management, data quality, data accuracy, and the way that the organisation manages that data. The tooling that you apply on top of that it’s important but it’s actually the data quality that that makes a massive difference to that model,” said Petty.
Reshaping the economic model of the internet
By getting AI right with the correct architecture and guard rails in place, Petty said in the coming years there will be a “fundamental change in the internet and the way we use it to serve customers and as a channel to market”.
He pointed to Apple’s recent announcement of installing Open AI’s ChatGPT LLM into its iPhones, and how other smartphone manufacturers have been pushing in this direction. This will change how people access the internet, as consumers will be using AI as their “primary agent” to the internet.
“Generative AI capabilities running on the device have the potential to fundamentally reshape the economic model of the internet that we have today.
“The economic model on the internet today is really simple. You go to a search box, you type something in and depending on who paid the most money or did the best search engine optimisation, you’re presented with a set of results that dictate what you do and from that is driven the whole advertising economic model that sits behind the internet,” Petty explained.
The executive highlighted how consumers can circle with a stylus an image on screen of a Samsung device, which can then produce purchasing options, as an example of a “different digital channel.”
In a potential world where people are circling images or talking to LLMs to replace the billions of daily ‘Google searches’, “I can’t use search engine optimisation to present Vodafone as the best experience. I need to interface into the LLM or the bot running on the device to present the experience with Vodafone is better than it is with other operators,” said Petty.
Lack of infrastructure in developing markets
The AI revolution and the fundamental landscape change it could deliver will trickle slowly down to emerging markets, as many expect.
Petty said availability of infrastructure such as data centres from hyper scalers to support any next generation technology is “not as well deployed”.
“We build more capabilities in our own data centres using the same technology but hosted in our models. We have very strong financial services business in Africa, with M-PESA and we’re applying gen AI capabilities to our financial services business, which we don’t have in Europe. So there are different areas and different focus points that we use [gen AI in].
He highlighted another challenge, one that has been a constant in emerging markets, and that’s availability of smartphones.
“Africa still has huge proportion of the population that don’t have smartphones and therefore chat is not an effective interface for them, they still use USSD as an interface. There’s a lot of differences in those [emerging] markets. How they apply the framework and the architecture [for AI] is still the same, but the use cases we build is different.”
Vodafone’s chief technology officer laid out the initial plan on operator group will be tapping into AI but warned to manage expectations as it is not the silver bullet operators are looking for, yet. But to truly harness AI and have meaningful impact to operations, Petty warned solid ground work is needed in governance, data and partnerships.
The TM Forum’s CEO Nik Willetts revealed recently that only a handful of operators have successfully used AI to affect their bottom-line, but even then, these would have been small gains, for now.
How AI will affect developing markets remains to be seen. But if investment into the technology continues at this rapid rate, new innovations to boost lower to middle income nations could be here before we know it.
