Windstream uses Azure OpenAI Service to empower employees and transform business outcomes

With the vision of creating more fluid workstreams and empowering its employees, Windstream turned to generative AI and the Azure OpenAI Service from Microsoft to revolutionize its operations.

Leveraging Generative AI

Russ Bartels, Vice President – OSS, Data & AI at Windstream, initially saw the potential of generative AI to enhance data science on top of the company's big data platform. In early 2023, Windstream entered discussions with Microsoft about utilizing Azure OpenAI Service for AI voice analytics. The goal? To extract valuable insights from customer interactions and call transcripts, leading to more efficient workflows and improved customer service.

Customer representatives play a pivotal role in analytics, particularly in topic modeling, summarization, and sentiment analytics for customer support and feedback analysis. They collect and preprocess data, providing valuable inputs for topic modeling, while also assisting with summarization through automated tools. Additionally, customer representatives utilize summarized information to understand customer concerns quickly. In sentiment analytics, they label data for model training, analyze feedback sentiments, and identify emerging trends. Their involvement ensures relevant and actionable insights from these techniques, enhancing customer satisfaction and organizational performance.

"Summarization allowed for less notetaking by our customer service representatives because AI can do it for us," said Bartels. "We haven't replaced our customer service representatives, but we're now a lot more successful in analytics and topic modeling, summarization, and sentiment analytics because we can leverage generative AI and its capabilities with our current AI voice analytics solution."

Outage Notifications: From Technical Data to Customer-Friendly Communication

With growing proficiency in prompt engineering, Windstream employed Azure OpenAI Service to create outage notifications. By analyzing technical information and error codes, the company transformed this data into customer-friendly messages to inform users about issues and expected resolution times. This seamless integration of generative AI allowed Windstream to turn complex technical data into easily understandable and actionable communications for their customers.

Internal Knowledge Management: Scaling with Azure AI

Windstream's next move was to utilize Azure AI, including cognitive search and OpenAI's Davinci, to index their vast amount of internal social media data and documents, which included approximately 100,000 indexed documents on their Confluence wiki. This indexed data was made available to Windstream's custom-built GPT (Generative Pre-trained Transformer) platform, hosted within Azure Kubernetes Service (AKS). Doing so allowed employees to access indexed knowledge and answer questions, leading to increased efficiency and more informed decision-making.

"We're leveraging Azure Cognitive Services, Cognitive Search, Azure OpenAI, and Davinci for our internal knowledge management solution," explained Bartels. "We've indexed about 100,000 documents, making it available through our custom-built GPT platform within AKS. It's scalable, performant, and ensures we provide accurate information to our users."

Embracing the Customer Feedback Loop

To ensure that the AI-driven knowledge base continues to improve, Windstream implemented a customer feedback loop within their GPT solution. This feedback mechanism allowed employees to provide positive or negative feedback on the answers provided by the AI. By analyzing this data and making continuous iterative improvements, Windstream aimed to fine-tune their AI models to provide more accurate and reliable answers over time. Bartels described the significance of this feedback loop in the context of empowering employees and creating efficiencies within the company: "We've built in the thumbs up, thumbs down in our internal GPT solution, so that employees can provide positive or negative feedback and we in turn can learn, grow, and change based on that.”

Expedited Customer Onboarding

A swift, self-sufficient onboarding process ensures that new employees receive the necessary training and information about the company's procedures, policies, and culture, which minimizes the chances of errors or misunderstandings that could lead to costly mistakes. Furthermore, an expedited onboarding process enables HR departments and managers to focus on other critical tasks, thereby streamlining their workflow and preventing bottlenecks during the recruitment process. 

Windstream recognized this challenge and utilized generative AI to assist new employees with questions related to Windstream policies, products, and services. "We're using generative AI to help new employees onboard,” Bartels said. “One of the greatest challenges is getting new employees to the point where they're self-sufficient, because self-sufficiency translates to more effective workstreams."

Scaling Up AI Adoption in the Field

Windstream's enthusiasm for generative AI and its numerous success stories encouraged further exploration and scaling up of AI adoption. The company envisioned using generative AI to enhance the capabilities of their field technicians, providing them with real-time assistance in troubleshooting technical issues.

According to Bartels, "One of our next big use cases is going to be using generative AI for our field technicians. We're going to feed a large number of technical documents into an index that's specifically designed with technicians in mind to allow our field teams to work through issues faster. Doing so improves the efficiency and responsiveness of customer care representatives and creates more cost-saving opportunities for the business."

Challenges and Learnings: The Journey Towards Excellence

As with any emerging technology, there were challenges and learnings during Windstream's AI journey. Fine-tuning the prompts and training the AI models for greater accuracy were crucial to leveraging generative AI effectively. The Windstream teams were committed to educating teams about prompt engineering.

"We've grown exponentially on prompt engineering over the last few months,” Bartels said. “There's a learning curve to it. As fast as this is moving, it is easier than it once was, but it still requires work and smart people to leverage it and turn it into something valuable."

The Road Ahead: Aspiring for AI Excellence

Looking ahead, Windstream aims to continue its partnership with Azure OpenAI Service and Microsoft to further enhance its AI capabilities. Bartels envisions a future where the company can fine-tune and train AI models like GPT-3.5 and GPT-4 on its data, leading to even more accurate and relevant responses. With the support of Microsoft's very secure Azure OpenAI Service, Windstream is confident in its ability to embrace new technologies and maintain its position as a pioneering force in the telecommunications industry.

"Being able to take the data, easily apply machine learning and allow anybody in the company to gain value from it is what I think is so exciting about generative AI," says Bartels. "It's made it easier for people to use and apply technology."