BT Group’s Digital unit accelerates AI use cases and navigates GenAI
Like most companies, telcos are grappling with what generative artificial intelligence (GenAI) means for their businesses as the hype continues to churn around tools like ChatGPT. The new tools are important advances and need to be studied, but there is a risk that they could become distractions from existing AI strategies and transformation programs. Dr Zoe Webster, AI Director for Data and AI Solutions at BT Group, talked to Inform about how the operator keeps GenAI in perspective as it focuses on accelerating the use of AI across the organization.
For Webster, GenAI is another set of tools in BT Group’s AI toolkit. “It can’t help us with everything, and we shouldn’t be using it for everything. It could be a sledgehammer to crack a nut [in some cases],” she said.
The rapid rise of GenAI has taken businesses by surprise and, in every sector, companies are scrambling to get to grips with it. BT Group has been busy with flurries of activity over the last few months to assess how the technology can be applied “safely and securely.”
“We’re cautiously excited about it, because we can see that there are risks as well as opportunities and things we need to think about very carefully,” said Webster, noting that BT Group has a head start in this area because its research division at Adastral Park has been working on some of technologies that underpin GenAI.
The new technology demands attention and resources, but it is important that this effort does not detract from ongoing AI and data projects. To avoid that, Webster said she is focused on “ensuring we don’t derail everything else that we’re doing,” and this necessity is recognized at the highest levels of the company.
“BT Group understands that [GenAI] is just the latest development and that we’ve got other work going on that will continue. We keep people focused on what is the business need and what is the insight you want from data… We can then decide what approach makes most sense,” she said, whether that is GenAI or other tools.
Joining up AI work at BT Group
Webster joined BT Group in November 2020 and was brought in to build up a “centre of enablement” for AI by Adrian Joseph, Chief Data and AI Officer at BT Group, who was appointed in February 2020 to lead the operator’s AI and data strategy. She now heads up a team of around 30 people who support the operator’s business units in leveraging data science.
“My job is to help to accelerate the use of AI and data science across the whole organisation…[and] make it easier to generate more value from our data. There is a lot of high-quality and focused data science and AI going on in pockets [of BT Group], but it has been subscale and not joined up. We provide a horizontal group that can draw those threads together, which makes it much easier for us to develop in one area and reuse in another,” she explained.
There are data scientists in many parts of the business who are specialists in their area with specific “domain knowledge.” They could be focused on a product set or a particular customer segment, for example. Members of Webster’s team will join them in squads to develop use cases, bringing more advanced AI expertise to the table.
Webster admits she was “quite choosy” when building up the team, most of whom were external hires to complement the experienced data scientists already in the business units. Along with wanting the group to be diverse and inclusive, she looked for “people who are just very curious” and “driven by the business needs and not taking things at face value, but really understanding and getting under the skin of what the need is,” she said.
BT Group has been working on AI since the mid-1990’s. In the last few years, it has become more central to the operator’s strategy as it looks to extract more value from its vast amount of data. Indeed, the target for the Digital unit at BT Group is to derive more than £500 million of internal value from AI over the next five years through “improved customer experience and enhanced efficiency” as part of the operator’s broader transformation.
Acting as an AI “hub” for the operator, her team works on standalone projects, supports other business teams, and shares learnings with the rest of the company.
At the same time, learnings from the domain expertise can be generalized and applied to other parts of the business.
Her team also works closely with AI researchers at Adastral Park and can help to bring getting new developments into the business more quickly.
“The magic really happens by bringing together the domain expertise and the AI expertise,” she said.
Revving the AI engine
Digital worked with partners Google Cloud and Datatonic to develop a machine-learning operations platform, called the AI Accelerator, that shortens the time it takes for AI use cases to be implemented. The operator set an ambitious goal to reduce that time from six months as of April 2022 to six days by the end of March 2023. It hasn’t hit that target yet, but the time has been dramatically reduced and is currently around 14 days.
“The challenge most companies find, and we’re not alone in this, is that it’s quite straightforward to generate a prototype or a demonstration of what AI can do. The difficulty is getting that into live production…Through the development of the AI Accelerator, that is becoming a lot easier for us to do,” she said.
Running on Google Cloud, and enabled by the operator’s data migration to the cloud environment, the platform saves time by providing templates for the administrative and technical processes required to launch AI use cases, such as compliance and testing. It will also monitor the performance of AI models.
“We want data scientists to focus on what they do best, which is finding insight from data. We don’t want them having to worry about the various processes to get their models into production,” she said.
AI in action
Most of Webster’s work is for internal AI use cases that fall into three broad categories: improving customer experience, supporting the generation of new and enhanced products and services, and increasing productivity in operations.
“We’re working on AI use cases that can generate value for the business by being able to do our processes faster, get more insight into our customers or networks, and respond more effectively,” she said.
As data science and ML evolves, talk of machines replacing humans has captured imaginations and sparked social and ethical debates. But people will remain essential for realising the best outcomes from AI, according to Webster.
“In the context of GenAI, as a new era of human-machine collaboration, the beauty comes where we harness the best of our creativity and ability to problem solve with the technology and the machines that are going to allow us to do things much more quickly or do some of the duller things for us,” she said.