Why this state-backed impact investor is pushing its startups to adopt Claude Code

AI coding tools have become one of the more talked-about shifts in how software gets built. Anthropic has carved out a clear lead over OpenAI in the coding space specifically, and Claude Code has generated billions in revenue since launch – a signal of how quickly the tool has been taken up across the industry.

For impact startups, Jonas Bergqvist, investment manager at Almi Invest Greentech believes the tool has a lot of potential – but that the window to get ahead won’t be open forever.

A person's tool, not just a developer's

His case for Claude Code isn't primarily about engineering capacity. He thinks the real leverage comes when people with product backgrounds – not just developers – get their hands on it.

"It makes it possible to move much faster from an idea and a customer need to something you can actually test," he tells Impact Loop.

Features that would otherwise sit on a roadmap for months can get built in days. Fewer people can ship more. For small teams operating with limited resources, that compounds quickly.

The small-team advantage cuts both ways, he adds. Large organisations face a messier transition.

"A rollout across hundreds of developers risks becoming much harder to manage," he says.

His broader point is that this isn't something to delegate. Bergqvist thinks leadership needs to be using these tools directly – not just hearing about them second-hand – to properly understand what they're capable of.

Meanwhile, one of Europe’s biggest tech unicorns is already using Claude Code to get ahead.

45 seconds instead of ten minutes

Voi, the European e-scooter operator, has been rolling out Claude Code in its operations in recent months.

Magnus Dahlbäck, Voi's head of AI, joined the company four and a half years ago to build out its analytics function – and describes the current AI push as a natural extension of that work.

"We want to apply AI wherever it makes sense. But we start simple and focus on internal efficiency – things we can control ourselves and learn from," he tells Impact Loop.

The clearest example is how Voi now handles parking fines. Many of the cities Voi operates in issue penalties when vehicles are parked incorrectly, and the process used to be entirely manual: a team working through scanned PDF documents, pulling out addresses, fine amounts and account numbers by hand. Each case took around ten minutes on average. It now takes 45 seconds using a solution the team built using Claude Code.

One person on Dahlbäck's team built the tool and put it into production in roughly two weeks. It reads incoming documents, extracts the relevant data, and surfaces it to a handler for review and sign-off.

Everything else is automated. Accuracy, he adds, is now higher than it was under the old manual process. The combined savings are hundreds of thousands of euros a year.

"The models keep getting better and better," he says.