Login
Register
The operator’s workforce is savvier about using data but access to customer information must be safe and controlled, says Mark O'Flaherty, Managing Director for Data & AI, Digital.
Data democratization is not a free-for-all for BT Group
When it comes to data democratization at BT Group, protecting customer data and using it responsibly are foundational. In an interview with Inform, Mark O'Flaherty, Managing Director for Data & AI, Digital, explained how BT Group’s new data architecture is built on a safety-first principle and why some parts of the organization are leveraging more data than others.
BT Group’s starting point for making data accessible is ensuring that “information can be accessed in a safe and controlled way … it’s not a free-for-all,” he said.
Once the safety rules are in place, BT Group provides toolsets, that enable people to explore the data to solve their business problems, along with access to its data scientists to guide the experiments and hypotheses to be tested on the data.
The operator’s data democratization efforts are in part helped by the proliferation of AI and a cultural shift within the workforce that is increasingly savvy about data. O'Flaherty has found employees to be much more knowledgeable about machine learning and data set manipulation and eager to grow their understanding of how to use data to drive business value.
“Some of our colleagues who have never done programming before are now using sidekick modules through Python to explore the data in their space…This is a significant growth compared to three years ago when we were not doing much of this. Now we are [looking] at how we can support our colleagues to do that better.”
O'Flaherty took over leadership for Data & AI at BT at the start of this year. While he is relatively new to the role, he has held senior IT positions since 2008, most recently serving as CIO for the Emergency Services Network in BT Group, and prior to that, CIO at BT Enterprise.
Modernization drives democratization
At the heart of BT Group’s data transformation is a strategic partnership with Google Cloud announced in March 2022 that encompasses data migration to the public cloud and leveraging the cloud provider’s AI and data analytics tools. BT Group has designed a data mesh architecture, which organizes data streams from various sources, as well as a single data fabric that applies the access controls for the entire business.
The data modernization program is complemented by composable IT components built on TM Forum’s Open Digital Architecture. BT Group achieved “Running on ODA” status in early 2024 in recognition of its work to simplify its architecture and establish “a robust, data-first, AI-ready infrastructure.”
The pace of change varies across the telco depending on the type of data and where it originates. Broadly, when the data ownership is “closer to home” and relates to networks, usage is more advanced than when customer data is involved.
“Where the data is less personal and more around the network, the infrastructure, the topology … we seem to be further ahead. Network engineering as part of our big mission to roll out fiber and 5G is core to what we do,” he said.
Part of the reason for the greater maturity with network data comes back to the operator’s priority to ensure it is “using customer data and insights for the right reasons”.
But that’s not to say that BT does not use customer and product data for insights. It does, and that’s where the data fabric plays a role in setting the governance controls so that the data is used safely.
For example, the operator can glean insight into customers’ overall service experience or spot opportunities to suggest new products or offers. And the data analysis is evolving to give the operator deeper understanding of customer needs or issues.
As BT Group is in the midst of its rationalization program to retire legacy systems and move to cloud-based applications, one of the challenges in creating data products is the variability of the data that is coming into the data mesh.
For example, data about how customers pay their bills is straightforward and “relatively solid”, he said. “But there are other streams like service and activities and everything else, which are more uncertain. So, you’ve got this mesh of all of these different streams, and then through that you’re creating your data products.”
Further, data might reside in multiple systems.
“We’ve got systems in two worlds, one where we were and one where we want to get to,” O’Flaherty explains. The difficulty lies in being able to explain the variability of the data to the data consumers who may be extracting the information to make decisions.
