DSB, Denmark’s national railway operator, serves over 195 million passengers each year, making exceptional customer service critical. Yet, historically, DSB relied heavily on manual processes and internal assumptions. To change that, they turned to AI-driven insights, bringing in Conversational Intelligence to better understand customer needs, identify pain points, and guide smarter decisions across the organisation.
From gut feeling to customer-driven decisions
Before implementing Puzzel's Conversational Intelligence, DSB gathered customer feedback through traditional surveys, limiting their understanding of real customer pain points. Internally structured processes added friction to the customer journey, leaving teams with limited visibility into what customers were truly experiencing.
“Five to ten years ago, we didn’t have many insights into what our customers were actually thinking,” said Søren Kristian Steffensen, Analytics Manager at DSB. “We based many decisions on gut feelings and internal assumptions.”
Implementing AI-powered insights with Conversational Intelligence
To address these challenges, DSB integrated Puzzel’s Conversational Intelligence, an AI-powered solution that automatically transcribes, tags, summarises, and analyses customer interactions.
“We quickly gained insights into our challenges, opportunities and future possibilities,” said Søren. “Now, we genuinely understand customer struggles. We've flipped our approach, aligning digital services directly with customer needs.”
Clear results and strong impact
Since implementing Conversational Intelligence in December 2023, DSB has seen strong improvements across key areas:
- Customer sentiment increased from 43.7% to 58.8%, exceeding their target of 54%. “Our goal for 2026 is 67.9%, and we’re currently on track,” Søren noted.
- Tagging accuracy jumped from 65% to 95–98%, delivering more reliable insights and saving around 400 hours annually through automation.
- Improved experience across channels, with conversation analysis highlighting key pain points, such as website navigation, leading to targeted improvements.
- Better visibility across digital and written interactions: DSB also extended Conversational Intelligence to analyse text-based interactions, identifying resolution rates and SLA performance across tickets.
- Deeper, data-driven insights: Conversational Intelligence effortlessly analyse large data volumes, clearly identifying key customer issues. For example, out of 38,231 analysed conversations, DSB learned precisely that 30% mentioned their website, highlighting the need to make improvements to their website.
Søren emphasised, “30% of customers mentioning our homepage is a red flag, it means customers tried to resolve issues independently before calling us. That insight alone is invaluable.”
Moving from cost centre to strategic advisor
Conversational Intelligence has helped DSB’s shift from reactive customer service to proactive problem-solving. “Insights no longer die at the frontline,” said Søren. “Now, we have actionable data available company-wide, influencing strategic decisions.”
With valuable insight into customer needs and behaviours, DSB’s leadership can prioritise more effectively, whether it’s improving web content or supporting new product launches. “We’ve moved from a traditional, cost-focused customer centre to becoming strategic advisors within the organisation,” Søren added. “That shift required strong change management, made possible by reliable, AI-driven data.”
Strengthening agent support and development
While technology is the enabler, the real transformation at DSB has been human. “Change is 90% people and 10% technology,” Søren said. “The big difference maker is that Conversational Intelligence reveals issues that weren’t visible to the naked eye before.”
By uncovering blind spots and delivering valuable insights, managers can now coach more effectively, while agents better understand how their work contributes to the customer experience. “It’s hard to have performance conversations based on gut feeling alone,” Søren said. “Managers can now use objective, data-backed insights to support agents more effectively.” The results are better coaching, improved conversations and more motivated teams.
“Conversational Intelligence gives us tools to understand precisely what agents do well and areas needing improvement, dramatically improving agent satisfaction and the overall dialogue quality,” he adds.
Building on success with AI-powered insight
While Conversational Intelligence is already delivering results, it’s just one part of a broader transformation at DSB.
“This is part of a much bigger transformation,” Søren explained. “By expanding these insights into other areas in the business, we can unlock even greater value - for our customers, our teams, and our strategy.”
Next on the agenda: closer integration with digital channels to enable faster resolutions, more proactive support, and a smoother experience for passengers.
“Expectations for AI-driven solutions like Conversational Intelligence are higher than ever. We're increasingly trusting AI, recognising its enormous potential in customer service.”
Five to ten years ago, we didn’t have many insights into what our customers were actually thinking. Now, with the help of Puzzel, we genuinely understand customer struggles. We've flipped our approach, aligning digital services directly with customer needs

The term itself refers to the process where agents categorise customer inquiries. To ease the process, many companies create lists of frequently mentioned topics that agents can choose from.
However, this list can be long, one inquiry sometimes contains overlapping issues, and it can be challenging for agents to remember everything said during a conversation. This means that manual call tagging is often inefficient, time-consuming, and error-prone.
In fact, manual call tagging typically takes 15 to 45 seconds per call. In addition, our analysis shows that agents miscategorise about 30% of their calls, which results in inaccurate data sets.
Therefore, automating call tagging saves time and reduces the likelihood of errors.