Eneco moves to Microsoft Copilot Studio to better support over 1.5 million customers

Eneco is a leading provider of sustainable energy in Belgium, serving over 1.5 million customers in consumer and businesses markets. The company specializes in wind energy, solar energy, and e-mobility and is the largest solar energy producer in Belgium with more than 400,000 solar panels.

Eneco runs multiple service channels including web, mail, telephone and live chat. However, growing call volumes had become increasingly difficult to manage. The system was already coming under strain as far back as 2022, when inbound traffic almost doubled during the COVID pandemic. Recently, when energy prices spiked due to the crisis in the Ukraine, volume spiked again. “We simply did not have enough agents to answer all customer inquiries,” says Ellen Van Caillie, head of conversational AI at Eneco.

Eneco developed an automated chatbot to ease the pressure on call center employees and provide a better customer experience. As it turned out, the chatbot did not perform as well as expected. As Van Caillie says, “The AI model was something of a black box for us, making it difficult to train and refine. As a result, the model did not perform as accurately as we had hoped.” The result was that a large number of simple customer enquiries continued to be routed to call center channels.

To address the problem, the company decided to move to a new platform - Microsoft Copilot Studio. Eneco was already using multiple Microsoft technologies which worked seamlessly with the platform. Better yet, the development team found that Copilot Studio enabled better control of the conversational experience. “Microsoft Copilot Studio has provided us with dramatically better guardrailing than our previous chatbot and enables better intent recognition and control of responses to customer questions,” says Van Caillie.
 

Faster development

Using Copilot Studio, Eneco replaced its chatbot with an AI “agent” – an advanced, AI assistant driven by the powerful capabilities of Azure AI. Leveraging the no-code, graphical interface in Copilot Studio, the Eneco team developed their agent in just 3 months – half the time needed to develop the company’s original chatbot.

With Copilot Studio, the team was able to migrate all questions and answers developed from the original chatbot into the new platform with minimal changes.

To enhance the conversational experience, Eneco combined the standard Natural Language Understanding model in Copilot Studio with the Conversational Language Understanding (CLU) model in Azure AI. This provided a number of benefits, including greater customization of the intent recognition model and complex entity extraction. More customization means that the new model can better understand a wider range of questions without having to ask a customer to rephrase.

If the AI agent is unsure of the request, the customer is transferred to a live chat with a customer service representative, together with an AI-generated summary of the conservation. Connectivity between the agent and other platforms, such as Eneco’s third-party live chat platform, is provided by Seamly, Eneco’s technology partner on the project.

With Azure AI Translator, the Seamly platform is also capable of translating everything in the conversation. This covers both the interfaces and the full conversation (including inputs from the customer, bot, agent). For example, if a question is submitted by a customer in German, it can be automatically translated for Dutch-typing agents in live chat. This greatly increases the flexibility and capacity of Eneco’s workforce and also increases the accessibility of the AI agent.