Spark is New Zealand’s largest telecommunications and digital services company with a stated purpose “To help all of New Zealand win big in a digital world.” Since 2014, Spark has evolved from a traditional telco provider to a digital services provider, including sports media, health, and the Internet of Things (IoT).
Spark New Zealand was one of the early adopters of artificial intelligence (AI) in the country, so I had many questions for Kallol Dutta, who has the intriguing title of “Tribe Lead- Data & Automation”, Spark New Zealand. More on the later in the article.
Achieving the single source of the truth
In 2014, Spark was facing significant competitive challenges. Technology was evolving quickly, legacy revenue lines were declining and increasing competition from digital natives such as Netflix and Spotify.
Spark embarked on a Data journey partnering with Infosys. The ambitious project involved re-engineering the legacy platforms and building a solid data foundation. Re-engineering helped consolidate multiple data systems into a unified system that acted as the single source of truth.
Cloud became a crucial part of Spark’s technology strategy, utilizing Microsoft Azure with ability to integrate other cloud providers . Multi-vendor selection of machine learning tools was Microsoft Azure ML, R and Pythonand data engineering tools like Data Bricks. Snowflake is the data warehouse platform. Infosys Cobalt cloud enterprise transformation services played a key role in orchestrating this ecosystem for success.
Building on the data foundation, Spark implemented AI and Automation at scale by scaling internal capability as well as leveraging its partnership with Infosys.
A faster operational cadence
In 2018, significant cultural changes were happening at Spark, which brings me back to Kallol’s intriguing title, “Tribe Lead – Data & Automation.” Spark was one of the first companies in the world to flip completely to “Agile.”
Implementing agile work practices company-wide meant moving forty percent of Spark’s employees into cross-functional “tribes,” comprising people from IT, networks, products, marketing, and digital. The agile transformation has since reached all parts of the organization.
The primary motivation for agile was to improve the customer experience, speed to market, and empower employees. Agile puts direct ownership and accountability with the tribe and increases employee engagement because people have richer jobs with a broader perspective on solving problems.
Agile is about removing layers in the organization resulting in a faster operational cadence. A multidisciplinary tribe is working together to provide an end-to-end capability as opposed to the traditional serial process with different narrow centers of functional expertise. Agile is focused on: “How do I work across this group to deliver the outcome?”
A new tribe called “data and automation” reflected the strategy and the ambitions for data analytics in the company. The data and automation tribe working with Infosys was the testing ground for new services for the organization. The formation of the new tribe elevated data and analytics from a reporting and compliance function to a role of deriving commercial value from data.
Algorithms that reduce churn
Spark utilized Robotic Process Automation (RPA) for automating repetitive tasks. RPA enables the building of bots which makes the process of actioning much faster. There aren’t enough programmers and RPA enables “citizen developers” without CS degrees to create programs. Cognitive AI is used for pattern detection and intelligent automation, for example, in retention and propensity models. Conversational AI uses natural language processing (NLP) to communicate and understand contexts such as chatbots and email automation
Like many telco’s, customer retention is a significant issue, so not surprisingly, this is where Spark found some early success with AI, building a churn model to address customer retention. The churn prediction model automatically tracks which campaigns will be more effective in retaining customers at risk of leaving Spark.
The solution spans data engineering, modeling, campaign creation, and monitoring lifecycle. The details on this blew my mind and during my research, I kiddingly suggested to Spark they license its AI model to US carriers as churn is brutal.
Spark has moved beyond the churn problem to focus on understanding customer propensities with machine learning models, such as understanding interest in a new device or plan change using deep customer insight of the customer and the household.
Spark is planning to apply a similar approach across other use cases, such as linking capacity management of the carrier network with customer experience so that sales and capacity management can be adjusted dynamically based on need.
In Spark’s annual report , data was a driver of strong performance which also demonstrates the leadership team’s commitment to data driven value creation. I’m very impressed with the sophistication and results.
Spark has learned that only two percent of customers are engaged anytime. However, the tendency has been to engage one hundred percent of customers all the time. AI has enabled Spark to engage with customers only when there is a need instead of all the time. This is coupled with desire to lead out on responsible AI initiatives – i.e., AI which is transparent, explainable and free from bias has changed the game in terms of how Spark deals with customer-centricity.
The end game now is to democratize the AI across the enterprise so that every part of Spark uses AI as the decision-making tool. The new data and automation tribe is the AI factory for the organization. The initial focus was on marketing and product. Future opportunities will extend to the network, fraud management, and channel optimization to serve SME and enterprise customers.
Note: Moor Insights & Strategy writers and editors may have contributed to this article.