Chatbot: How Norwegian companies adopt AI without getting burned 

Most companies have long understood that they need to test and adopt AI. But how do you do that without wasting time, money, or losing control over your own data? Those who succeed start gradually and smartly.

Frøy started with the chatbot Frøydis 

Frøy is Norway’s largest expertise environment in aquaservice, a total supplier of fully integrated and sustainable vessel services. They also want to free up resources with practical use of AI.

– Frøy is a large company with many employees, spread along the Norwegian coast. We have grown over time with long and secure organic development, combined with targeted acquisitions and mergers. In all of that, it can become a jungle of information, says Per-Andreas Drevvatne, Head of DigIT at Frøy.

A single point of contact for employees 

In collaboration, Frøy and Computas have therefore developed Frøydis – an internal AI chatbot that helps employees find the right information quickly and securely.

– Those who adopt AI early learn faster and get better returns. It’s as much about the organization’s pure adaptability as it is about technology, says Pål Vermund Knudsen, Chief Architect at Computas.

– Those who dare to try gain insight into how AI works. They learn where their information is lacking, and they improve it, he says.

Frøydis searches internal sources such as the employee handbook, SharePoint, and other relevant systems. It delivers precise, source-referenced answers – and both usage and effect are visualized in its own dashboard.

– The goal is for Frøydis to be a single point of contact for an employee who has questions. Instead of turning to a colleague or immediate supervisor, one can perhaps start with Frøydis, says Per-Andreas Drevvatne.

– The goal is for Frøydis to be a single point of contact for an employee who has questions.

Per-Andreas Drevvatne, Head of DigIT in Frøy.

The checklist: Are you ready for AI on your own data? 

Computas often uses a simple checklist to clarify whether a company has a good basis for building its own AI solutions – a kind of maturity test:

• Do you have a concrete need or a role that can be streamlined with AI – and internal ownership of it?
• Do you have relevant data sources that can be trained on or linked to – and good enough data quality?
• Do you have the expertise or partners who can assist with implementation, responsibility, and management?
• Do you have security requirements or regulatory considerations that affect how the solution should be built?
• Do you trust that users will actually adopt the solution and provide valuable feedback?

If the answer is yes to most of these points, you are probably ready to build your own AI – safely and efficiently.

An AI assistant must be treated like a new hire 

A sustainable AI strategy is about secure solutions, transparency for employees and customers about how AI is used – and not least: full control over data.

– It’s like a new colleague you train: it must have access to relevant information, learn the language – and receive feedback. Gradually and controlled. The nice thing is that you also learn something back. Many customers discover where they have poor documentation and how it can be improved, says Pål Vermund Knudsen, Chief Architect at Computas. 

This ‘feedback loop’ is often an unexpected gain – you learn which questions are repeated, which systems are actually used, and what employees or customers are asking for that is difficult to find.

From experiment to value – quickly and safely 

As a strategic advisor, Computas helps companies and public enterprises map real needs, assess risk, and choose the right paths forward. The approach is about getting started quickly without unnecessary risk. The process begins with mapping needs and opportunities, before building a pilot solution that can be tested and improved based on actual use.

– We use ready-made building blocks, but stitch them together and adapt them to the customer’s needs, explains Knudsen. 

Many customers experience that they can see concrete results already in the pilot phase – and then scale the solutions to more departments or areas of use as trust and competence are built up. 

– Frøydis does not have access to all our data, we have consciously controlled the information flow from the start, by gradually feeding it with the sources we trust. Version 1 of Frøydis will have some weaknesses and limited source access, but together we are developing versions 2, 3, 4, 5, says Per-Andreas Drevvatne at Frøy.

– We have set up secure chatbots for customers in just a few weeks. When people see it in use, it gives a wow effect.

Pål Vermund Knudsen, Chief Architect at Computas.

Learning is the same as success 

It’s therefore not about revolutionizing everything at once, but about getting to know the technology through practical use of your own data and experiencing your own challenges.

– We have set up secure chatbots for customers in just a few weeks. When people see it in use, it gives a wow effect, says Pål Vermund Knudsen at Computas. 

– There are still many who are waiting for the perfect solution. But then you learn nothing, says Knudsen. 

The real danger in 2025 is doing nothing. Then you risk working as if it were 2015.