Betsson has compressed development cycles from 80 days to five after a company-wide AI push prompted one team to fundamentally rethink how it works.
The breakthrough came after CTO Fredrik Ögren outlined Betsson’s AI investment priorities during an all-hands meeting, sparking a transformation far beyond the introduction of new tools.
The team responsible for retention and engagement systems, including complex bonus and tournament mechanics, had traditionally operated through backlog refinements, planning cycles, and incremental releases stretching close to three months.
After the AI push was announced, the team broke into smaller units and used AI to help redesign the entire development process, not just automate isolated tasks.
“Once you see the value, it naturally becomes part of how you work going forward,” said Ögren. “You cannot say it was a great experiment and return to 80-day cycles.”
“When the mindset changes, it stays changed. That’s the shift we are seeing at the moment. It’s the next way of working,” he added.
Betsson had long invested in machine learning and analytics, but volume of data alone was no longer sufficient to drive the outcomes the company needed.
This prompted a structural shift, including the appointment of Cleber de Lima as director of data and AI, tasked with identifying gaps and building a genuine AI strategy from the ground up.
“I think that the most important part for us was to approach AI in a very pragmatic way,” Ögren said. “We didn’t want AI to become a monster within the company, without the right structure to manage outcomes effectively.”
Governance became a critical pillar of the strategy, with data control and responsibility treated as a foundation rather than an afterthought as the company scaled its AI capabilities.
“Structuring this is always a challenge, but two years in, we can see that without it, it would have been significantly harder to scale effectively,” Ögren said. “AI multiplies everything, and if it isn’t done right from the start, it just multiplies chaos.”
Betsson established a central AI Centre of Excellence, which mapped capabilities across the organisation and produced an internal maturity index to identify where intervention would have the most impact.
Structured programmes followed, covering agentic automation, AI-driven reporting models, and integration of AI-enabled features within internal platforms, each tied to a defined organisational need.
“The key was structure,” de Lima said. “We mapped the needs, then built programmes to elevate each area.”
Adoption was not without friction, as Betsson’s young workforce began independently downloading tools and setting up applications, creating an unplanned wave of unsanctioned AI usage across the business.
“We started seeing people downloading tools or setting up applications on their own. Not with any bad intent, just curiosity,” Ögren explained, noting that information security teams were brought in to assess the scale of rogue usage.
“We still encourage people to experiment, but with responsibility,” de Lima added. “Then we take what works and put it into production. At that point, it becomes a proven methodology that others can follow.”
On the question of building versus buying, Betsson draws a clear line between core business capabilities and more standardised operational functions handled by third-party solutions.
“Anything that sits close to our core business, we believe we need to build ourselves. That’s where we create differentiation and real intellectual property,” de Lima said.
“If you want to do something truly advanced with your own systems, something that creates value for your customers, then you need to build it yourself,” Ögren added.
Human accountability remains non-negotiable throughout Betsson’s AI model, with every AI-generated output passing through quality gates and a human reviewer before it is released.
“AI augments cognitive capacity but cannot be held accountable. Every AI-generated output passes through quality gates and a human in the loop,” de Lima said. “The developer releasing it owns it entirely and is responsible for it.”
As AI accelerates code generation, the bottleneck has shifted from writing to reviewing, prompting Betsson to begin integrating AI into verification steps as well.
“Now that code has been shortened, we can review the entire ecosystem and ask, ‘Okay, what can we do differently?’ We are not binding ourselves to any traditional methodologies,” de Lima explained.
Training and trust proved to be the biggest cultural challenges, with many developers initially sceptical of AI outputs due to earlier, less mature models or ineffective usage habits.
After a structured training rollout supported by real-world showcases, 97% of programme participants began actively using the tools, with adoption rising sharply following the intervention.
De Lima described spending a full day working directly with a senior developer on a real task: “By the end, he had a Eureka moment. He realised he could trust the tool because we were using it differently.”
Ögren cautioned against chasing every new tool without strategic purpose, warning that companies who moved quickly without structure now face serious challenges in scaling and maintaining what they built.
“Some companies moved quickly but are now facing challenges in scaling and maintaining those solutions. People built AI projects on their own laptops and then left, so who is running them now?” he said.
“We design systems to be reusable and scalable so they can evolve with the technology,” de Lima added. “AI will be central to our operations, but only if we use it deliberately, in ways that improve efficiency, revenue, or cost savings.”
