Buying AI tools is easy. Getting real value from them is something entirely different.
Cursor and Claude can have a major impact when used correctly. That is why Cefalo has established a structured training program that makes developers more confident, efficient, and aware of both quality and costs — enabling them to deliver true value to clients.
For Cefalo, AI is not just about adopting new tools, but about ensuring that the technology is used properly, efficiently, and responsibly. The result is a better workflow for developers — and more thoroughly executed work for clients.
Cefalo sees AI training as a necessary investment, both for its own competitiveness and for the clients the company works with.
In recent months, Cefalo has rolled out internal training programs in AI-supported developer tools such as Cursor and Claude, in addition to other relevant solutions like GitHub Copilot and similar technologies.
Around 70 developers participated in the first round of training.
– We realized early on that it’s not enough to simply give developers access to an AI tool, Handegård explains.
The experience was that many used the tools mainly as autocompletion, without adopting more advanced and agent-based ways of working.
– Some developers figure everything out on their own, but many continue to work as before, just with a new interface. That means you’re not leveraging the full potential of the technology. We wanted to do something about that, he elaborates.
Another important aspect of AI tools is the cost and how to use them responsibly. Many solutions are based on external LLMs and are priced according to usage.
– If the tools are not used correctly, it can become surprisingly expensive, says Handegård. – We had some examples of developers spending significant amounts in a short period of time without being aware of it. That’s why training is absolutely essential — for quality, cost control, and confidence.
This became an important driver for making AI training a structured and mandatory initiative, rather than something each individual developer had to figure out on their own.
The training is structured into several phases, with a clear focus on practical use:
1. Guided training
Experienced, internal users demonstrate specific cases. Participants work in parallel on their own machines and test the methods in real time, with close professional support.
2. Hackathon-like session
Developers are given specific tasks where they must apply what they have learned in practice.
3. Team follow-up
After one to two weeks, each team is followed up to assess how the learning has been implemented in their daily work, and to provide further guidance where needed.
– We wanted to make this as practical as possible, says Probal Sikder, Engineering Manager at Cefalo, who has been closely involved in the implementation. – Developers should work on their own projects, their own challenges — and be able to bring something directly back into their daily routines.
Feedback from developers has been very positive. Several teams report a clear shift in how they use AI in development work.
– Previously, AI tools were used quite superficially. Now we see that teams are working in a more structured and agent-based way. They let the tools handle more of the hard coding, giving developers more time to focus on complex problems, says Handegård.
For clients, this means increased productivity and higher delivery capacity.
– In practice, it means developers are able to get more done in the same amount of time, says Sikder. – Most of our clients don’t have too little to do — they have long backlogs. When we can help them get more done within the same framework, it is perceived as significant value to them, he concludes.