In our very first AI experiments, we quickly realized just how critical metadata and data quality are for AI
The spring sun lights up a modern office building in Ruoholahti, Helsinki, where one of Finland’s largest energy companies, Helen, has recently relocated its headquarters. Owned by the City of Helsinki, Helen is the market leader in retail electricity sales in Finland and, through its subsidiary Helen Electricity Network Ltd, is responsible for the capital’s electricity infrastructure. In recent years, Helen has taken significant steps toward the clean energy transition and aims to phase out combustion‑based energy production entirely by 2040.
Alongside this clean transition, Helen is undergoing a major digital transformation. At the center of this transformation is Mikko Muurinen, who leads the development of AI, data, and digital optimization at the company. “Heating is increasingly being produced with electricity, and the demand for electricity generation continues to grow. We need to prepare for increasingly weather‑dependent and variable electricity production, while ensuring sufficient supply during fluctuations. All of this has effectively turned us into a data company,” Muurinen explains.
Helen’s investments in data and AI help the company respond to this new reality. “The better we can predict what will happen in the next fifteen minutes, hour, day, week, or year, the better results we can achieve,” says Muurinen. Data also benefits Helen’s customers. “For example, we can offer products that better suit our customers’ needs or guide consumption to more optimal times.” In addition to supporting the business and customers, data also improves everyday work at Helen. “We are constantly thinking about how we can use data, AI, and analytics to make work smoother and bring AI assistants into different roles across the organization,” Muurinen adds.
At an early stage, Helen recognized that robust data management is essential if data is to be truly utilized – for example, in AI applications. “When you have vast amounts of data, you also need strong governance models to manage it,” says Muurinen. To develop its data management capabilities, Helen found a skilled partner in Sparta Consulting, and the collaboration began in 2020.
In the early phase of the partnership, the focus was on establishing data management practices at Helen. Sparta’s experts supported Helen in areas such as implementing a data catalog, as well as developing data quality and related processes. “We built capabilities for data management – defining what needs to be in place before you can even start effectively working with data,” Muurinen describes.
The documentation of data and the quality of metadata – the background information and context of data – were seen as particularly critical, as they form the foundation for using AI in business. “In our first experiments with generative AI, we clearly saw how important metadata and data quality are for generative AI as well. AI acts as an ‘amplifier’: it highlights what already exists in the data. The fundamental quality of the data must therefore be in good shape so that AI has high‑quality fuel to work with,” Muurinen explains.
AI also proved to be a solution for producing metadata. “Metadata creation often receives too little attention, even though it is a crucial part of improving data quality. Together with Sparta, we started exploring how AI could help overcome the ‘blank page problem’ and give data owners the tools to enrich metadata,” says Muurinen. The AI solution was initially developed as a pilot. In spring 2025, when Sparta became part of Efima, the project also benefited from Efima’s AI expertise. Today, the jointly developed solution is in production, supporting Helen’s data specialists in their daily work. “We are extremely satisfied with the collaboration with Sparta and Efima. Sparta has strong expertise in data management, which has been highly valuable in developing AI solutions,” Muurinen says.
Using AI to enrich metadata and improve data quality is only the first step on Helen’s broader AI journey. “We are building a foundation for future ‘chat with your data’ solutions. They simply won’t work unless the AI has sufficient metadata underpinning its operation,” Muurinen notes. According to him, AI will enable capabilities that were previously out of reach. “Decision‑making cycles are accelerating, and there is a growing demand for real‑time insights to support decisions. Our customers also expect better and more up‑to‑date services,” he continues.
On a larger scale, Muurinen believes AI is democratizing data work. “Data is increasingly accessible to everyone, and we need to provide ways to use it. We can no longer rely solely on dashboards created by someone else – the way we interact with data is becoming more conversational.” He also believes in agentic processes. “In the future, people won’t use data and AI to manually execute processes. Instead, there will be a swarm of agents operating in the background, surfacing insights when needed. Humans will guide the AI, while the AI executes processes and uses data.”
Muurinen encourages others working with data and AI to remember that even small improvements matter. “You don’t need to dive into the deep end with AI straight away. If AI can handle even 20%, the remaining 80% can still be left to people for now – and the balance can be improved gradually,” he says. “Working with AI is continuous development: improving one percentage point today and another tomorrow. If we can do something quickly that delivers value next week, that’s already a win. There’s no need to wait for a perfectly finished solution.”
Could we help you too?
You can contact us via the form below.