Six Nordic examples of AI in practice

Challenge

Clinical study reports can be extensive and require contributions from statisticians, scientists and technical writers across a long process.

Solution

NovoScribe combines generative AI, retrieval-augmented generation and domain-approved content. Anthropic and MongoDB describe it as a customer example.

Company-reported results

• MongoDB reports a clinical study report moving from around 12 weeks to around 10 minutes.
• Anthropic reports 90% less report-writing time. This is separate from end-to-end cycle time.

Goals and expectations

MongoDB says NovoScribe handled around 30% of reports at the time, with an expectation of more than 90% by year-end.

Challenge

Manual price calculation for enquiries took time and needed to be performed consistently.

Solution

The company implemented a machine-learning price-calculation algorithm using historical calculations to suggest a price draft.

Company-reported results

• Time per calculation fell by nearly 90%.
• Testing found a hit rate of at least 90%, meaning the model price often matched the company’s own calculation.
• At source time, it generated price drafts for around 20% of enquiries.

Challenge

Employees needed to find information in complex ERP data, and some enquiries took around 20 minutes to answer.

Solution

The company put a user interface on top of an AI integration with its ERP system so relevant information became easier to find in day-to-day work.

Company-reported results

• Response time is described as around 20 minutes to around 20 seconds.
• DI quotes that a first workable solution was built in around one and a half months.

Goals and expectations

DI reports an expectation that employees can be trained in around three months rather than up to two years. This is not a measured result.

Challenge

Customer service needed availability outside regular hours and a way to handle both simple and more complex questions.

Solution

Lunar launched a GenAI voice assistant in beta for selected customers to collect feedback and refine the service. 24/7 availability and human handover are designed properties.

Status and goals

• Lunar positioned the launch itself as “a first in European banking”.
• Lunar expected the assistant to handle around 75% of customer calls over time. It was not a measured result at source time.
• Tech.eu wrote on 24 October 2024 that Lunar was not hiring as fast after introduction, without job cuts at that time.

Challenge

Support needed to meet growing needs. Knowledge-base articles were long and lacked summaries and metadata that made information easier for the agent to find.

Solution

Mark uses a structured Salesforce knowledge base for service and can hand a human the conversation history, related articles and attempted solutions.

Company-reported results

• 37% of support cases resolved autonomously.
• 32,000 customer conversations since launch.
• 21% of customers voluntarily chose Mark over a human.
• NPS on par with human agents.

Challenge

Customer service needed availability across time zones while co-workers needed more time for higher-value conversations.

Solution

Billie handles simpler enquiries around the clock. Ingka says co-workers were reskilled in remote interior design, digital retail sales and related work.

Documented results

• From 2021 to 2023, Billie resolved around 47% of customer enquiries received.
• Ingka says this equals 3.2 million resolved interactions and nearly EUR 13 million in savings.
• 8,500 call-centre co-workers were reskilled.
• Sales through all Ingka remote customer meeting points reached EUR 1.3 billion at FY22 end.

Goals and expectations

Ingka wrote that it aimed to grow remote selling’s total-sales share to 10% in the following years.