We are back at Nextory’s colourful office, and this time co-hosted with Omni, Nextory‘s AI analytics provider, Data & AI Stockholm’s main community sponsor, and the proud sponsor of the upcoming Data & AI Stockholm Summit on October 14th.
The Omni team flew in from across the ocean to share their work with the community in person, and that commitment showed in every part of the evening. A case study followed by a hands-on workshop, food, drinks, and good conversations. A full night.
One Metric, Too Many Answers
Carl Niblaeus, Senior Data Engineer at Nextory, opened with a question the room recognised immediately.
Why is this dashboard showing X when my spreadsheet shows Y? Which should I use?
Nextory runs a subscription service for audiobooks and e-books across 10 European markets. Carl Niblaeus has been there since 2020 and came to talk about an 18 month platform rebuild that started with a familiar problem.
Ungoverned dashboards, spreadsheets nobody owned, definitions that had drifted over time. Carl called it the dark matter of the platform, and coming from someone who did his PhD researching actual dark matter, the metaphor landed well.
The clearest example was churn. One metric, four tables, four different numbers. Nobody had decided what it actually meant, so nobody could say which one was right. For a person that erodes trust. For an agent it is worse, it just picks one.
How They Fixed It
The answer was not a new tool. It was agreeing on what things mean, and then building a platform around those agreements.
Over roughly 18 months, a small team rebuilt the whole thing. The approach was to create a clear separation between what the data sources say, what the business means, and what different consumers need. Each layer has its own specification, and the SQL that runs in production is generated from those specifications rather than written by hand.
The layer that took the most effort was the business layer. Talking to people across the organisation, earning buy-in, finding definitions that everyone could live with. That is where most of the time went, and Carl was clear that it does not end.
Get this layer right and downstream follows. It is a program, not a project.
Carl shared the results of an evaluation he ran, putting the same business questions to both the old and new platforms using an AI model. The old platform returned several different values for the same metric. The new platform returned one. Where values still differed, they reflected genuinely different definitions, and a more precise question resolved them.
The churn metric that once had four answers now has one.
Where They Are Now
The platform shift changed more than the numbers. Across the business, people who previously had no way to get answers without asking the data team were now exploring on their own. The shared language around data that the rebuild created started showing up in meetings and everyday conversations, which is probably the clearest sign that something actually worked.
But Carl was also honest about what keeps them up at night. Agents combine data in ways nobody reviews, and the model has to carry the correctness. New metrics need to land in the governed model or you risk ending up with a second version of something you already have. And like most teams going all in on AI-powered self-service, the cost model is still evolving.
Where They Are Now
The platform shift changed more than the numbers. Across the business, people who previously had no way to get answers without asking the data team were now exploring on their own. The shared language around data that the rebuild created started showing up in meetings and everyday conversations, which is probably the clearest sign that something actually worked.
But Carl was also honest about what keeps them up at night. Agents combine data in ways nobody reviews, and the model has to carry the correctness. New metrics need to land in the governed model or you risk ending up with a second version of something you already have. And like most teams going all in on AI-powered self-service, the cost model is still evolving.
Then Omni Took Over
The second half of the evening belonged to Omni. Their platform sits on a semantic layer, which means when you ask it a question in plain language, the answer comes back grounded in governed definitions rather than a fresh interpretation of raw data. SQL, spreadsheet formulas, natural language, it all runs through the same layer and returns consistent results. With the right prompting, you can go from a question to a dashboard or a full analysis without writing any code.
The workshop used Stockholm Airbnb data as the playground. People split off and started exploring, and it got creative quickly. Some built dashboards. Others took the data in directions nobody had planned for, finding patterns in neighbourhoods, building out scenarios, seeing what the tool would do if they pushed it. The results genuinely surprised people.
And What’s Next?
The workshop results were impressive. Things that used to take time were happening in minutes, and people who would not normally touch a data tool were getting answers on their own.
And seeing that right there at your fingertips, well, it made a lot of us ask that million dollar question.
Are we going to lose our jobs?
Those are exactly the conversations worth having, and if you want to have them with the people closest to where this is all heading, October 14th is a good place to be.
The Data & AI Stockholm Summit brings together practitioners, researchers, founders and leaders from across the Nordic ecosystem for a full day. Industry leaders, real implementations, and the people shaping what comes next. To stay up to date with speakers and agenda, check the official Summit page.
Thank you to Nextory for hosting, to Omni for making the workshop possible and for making the trip to be with the community, and to Carl for a talk that left the room with plenty to take back.
Mark your calendars, we have unfinished business on October 14th!
Data & AI Stockholm 💙
Stay Connected
If you’d like to stay connected with the community and hear about future events, the best places to find us are: