Digia’s Hammer programme: turning AI experiments into actionable insight

Digia has launched an internal AI development initiative called Hammer. Its goal is to enhance the use of AI in everyday work and to develop innovative methods for leveraging AI capabilities in software development. A mob programming session in Jyväskylä was a great example of how Digia’s people are making AI a tangible part of the coding process.

In October, Matti Saarikallio, Senior Manager and leader of Hammer’s Software Development stream, brought together a group of software developers for a mob programming session, in the spirit of Digia’s continuous learning. This time, the challenge was to build an electricity price forecast using the GitHub Copilot AI agent.

Coding together and sharing knowledge

Mob programming is a practical approach to collaborative learning. The idea is to code as a group on one machine, taking turns to write the code. While this might sound daunting, it creates a safe environment where everyone receives support and guidance from the other participants. When learning AI (as with many things), hands-on experience is essential rather than just studying theory, as emphasized by the session’s experienced leaders.

“When we do mob programming, we figure things out together. The organisation’s understanding grows when information is sought and used collectively,” summarises Anssi Ylönen, Group Manager.

The role of AI in software development

Saarikallio and Ylönen have reflected extensively on the impact of AI on software development. Their current view is that the role of a coder will expand, and the working environment may become even more complex. In the future, a coder’s tasks will likely involve more specification work, as AI can automatically generate code based on those specifications. Mob programming is a great way to spark discussion about what can and cannot be done with AI.

“The most surprising thing about experimenting with AI has been how well it produces code from commands. On the other hand, quality and maintainability remain concerns, and AI adds complexity to the overall solution. The fact is that many things are likely to change. It will also be fascinating to see how AI influences recruitment criteria in the future,” Ylönen reflects.

What did the joint coding session teach us?

During the mob programming session, the team found that the framework they used worked well: specs → product requirements → technical specification → implementation plan.

Development progressed smoothly, and a near production-ready solution was achieved in just a few hours. However, quality and testing were key topics of discussion. The question of how to ensure the solution’s security also arose.

“It’s good to think about checkpoints in advance to keep track of what you’re doing. Setting milestones is important. When we know where we’re heading, things stay on track. This time, we also had senior colleagues present who definitely understood what was happening in the code. But you have to be careful with AI, as I recognise the temptation to press ‘yes’ and move on without reading the code,” Saarikallio summarises.

Hammer will keep striking in many ways

AI has become an integral part of everyday life. Through experiments, projects, and learning, it has earned a permanent place in Digia’s toolkit. The work to increase know-how under the Hammer project will continue – and likely accelerate.

Mob programming events are a concrete example of how AI expertise is increased, developed, and shared at Digia – with a twinkle in their eyes, they hammer away. That’s why the next session will tackle a challenge using Claude AI, for comparison.

Wishing everyone successful and insightful AI experiments!