Ruukki: How to make customer feedback more useful with generative AI

How to make customer feedback more useful with generative AI

Manufacturing

Ruukki Construction Oy is a leading Finnish manufacturer of steel-based construction products as well as a service provider. Ruukki conducts a yearly net promoter score (NPS) survey in 10 countries where they receive a lot of open feedback in numerous languages. Ruukki turned to Vincit to find a way to use generative AI to quickly and easily summarize this feedback and present it in a more useful format.

The challenge – open feedback in many languages

As part of the NPS survey, customers give feedback and improvement suggestions. Ruukki’s challenge has been to analyze and draw conclusions from this vast amount of data in different languages. It took hours of work from their research partner’s analysts to prepare summaries in previous years. This time around, Rukki wanted to test what modern AI tools could do with their feedback data.

This AI-supported analysis provided us with an amazingly accurate overall view of our customer sentiment. And more importantly, the summary and the recommendations help us to choose the areas that need improvement. That’s what really matters – to understand our customers’ needs and offer the best customer experience.

Pia Salonen, Development manager / Ruukki

The solution – using generative AI

Initially, Vincit proposed that Ruukki could use generative AI for summarizing their feedback. To explore the feasibility of this approach, Vincit conducted a proof of concept during a weekend hackathon, using data from Ruukki’s NPS survey. This experiment demonstrated that generative AI could accurately and effectively translate and summarize a large volume of information.

In the proof of concept, the idea was to translate all the feedback into English and analyze each piece to determine its sentiment (positive, neutral, negative), identify up to three keywords, and assign a topic. Additionally, the translated feedback was organized by country and business. Instructions were also provided on how to analyze the feedback comprehensively, summarize the positive and negative aspects, and offer suggestions for improvement.

Following the success of the proof of concept, the next step was to summarize the data from Ruukki's most recent NPS survey at Ruukki. The Vincit team used generative AI to translate the data into English, categorize the feedback into predefined categories, and extract insights related to three key questions:

what are Ruukki’s top 3 strengths?
what are the top 3 topics that customers refer to as improvement areas?
what are the possible recommendations for improvement actions?

In addition, the data needed to be handled in a safe and secure manner using Azure OpenAI API in an isolated environment where no third party could access the information.

Expertise used

Generative AI (LLM, Sentiment Analysis)

Development and API integrations

Agile way of working

Data handling and preprocessing

Security and compliance

Validation and mitigation of GenAI hallucination and bias

The impact – useful sentiment analysis and feedback summaries

Saving time by making data easy to digest

Generative AI translates and summarizes the NPS survey data to answer three key questions for Ruukki – saving a lot of manual work.

Helping Ruukki to offer a better customer experience

With easy-to-analyze results from the NPS survey, Ruukki can better understand customer needs and where to focus their efforts on further improving their products and services.