By Gavin Mitchell, Altice USA; Alex Gottwald, Palantir Technologies; Austin Atmaja, Palantir Technologies; Bruce Gatete, Palantir Technologies; Shane McWilliams, Palantir Technologies; Kate Van Horn, Palantir Technologies
Preventable service visits pose a significant burden to operators in the form of financial costs to the business, misallocated technician time, and degraded customer experience. Altice USA and Palantir set out to take a user and data-driven approach to solving this problem, leveraging the latest advancements in machine learning, generative AI, and data modeling.
Altice USA and Palantir built a network and customer model, replicating real-world business workflows in a digital twin. This model utilizes data from an array of sources including physical network topologyfor both Hybrid Fiber-Coax (HFC) as well as Fiber-to-the-Home (FTTH) networks, customer and devicebased service telemetry, customer contact points and technician service visits. Within the Palantir platform, this data is then leveraged by GenAI and LLM tools to elucidate trends in flows, which allows human operators to enhance their analysis and strategies.
Within 3 months, the initiative team brought new troubleshooting models to production by integrating recommendations within customer care tools indicating to the care operator as to whether a service visit was necessary based on the recent experiences for the customer.
In this paper we will walk through the data collection, modeling and implementation work completed by the team to take the ideas from conception to production. We will also share some of the early findings from the A/B testing which have identified a 7% reduction of preventable service truck rolls compared to a control group and 8% reduction in Average Handle Time (AHT) for care agents. Both represent an improvement in the operations of the business, the experience for our customers and employees. Next steps for further developing models and integration into operational processes will also be discussed.