Scaling AI across 160 independent business units on AWS with Visma

Visma has been at the forefront of technological innovation since its founding in 1996. Through a hypergrowth business strategy, the company has expanded to more than 160 decentralized, highly autonomous business units, each with their own tech stack and product development goals.

For Visma, giving its software and engineering teams agency over their product lifecycles is key to remaining competitive in the long term. “We want to make sure that everyone is on board and making progress with AI,” says Alexander Lystad, chief technology officer at Visma. “In setting standards and benchmarks for our product owners, we still want to keep our business units in the driver’s seat.”

Visma’s developer-first culture meant that many business units were already experimenting with AI technology. For example, several teams had adopted Amazon Bedrock to build generative AI applications and agents at production scale. “It’s super convenient that our engineers can consume a wide variety of models, with the kind of reliability and quality we’re used to, through Amazon Bedrock,” says Lystad. “As a developer, you don’t need to learn something different from AWS. It’s all set up within the interface and frameworks you know.”

Building on its prior successes, Visma collaborated with the AWS team to craft an AI strategy that prioritizes developer autonomy without compromising its regulatory compliance or data security. To drive adoption across the entire organization, Visma’s strategy consists of four strategic pillars.

The first pillar is AI-native product development. “We can build a lot faster and create better products by taking advantage of AI in both the engineering and product functions,” says Lystad. Using generative AI, Visma’s software teams have rapidly modernized their old code bases, helping them switch from commercially licensed databases to cost-effective, open-source technologies.

Experiencing the benefits of AI technology firsthand, Visma’s teams wanted to pass on those efficiencies to customers. This led the company to make AI-native products the second pillar of its strategy. By delivering new features to its customers, Visma helps them automate manual tasks and extract business intelligence from their operational data.

For example, Visma deployed a multi-agent system that is trained on tax and legal domains for the German market. One of Visma’s solutions, Taxy, has processed over 10 million tax cases, while its latest agentic AI solution recently demonstrated its deep domain quality by passing the rigorous German Tax Advisor Examination. “Driving automation has been a core aspect of Visma’s product line, and with AI, we can do more of it,” says Lystad.

Visma also saw an opportunity to empower its sales and marketing teams. The company’s third pillar, AI native-growth functions, focuses on automating go-to-market workflows. By providing its teams with AI tools, the company can deliver hyper-personalized experiences and resolve customer support cases more efficiently, improving lead generation and revenue growth.

Lastly, Visma has focused on building an AI-native workforce through internal restructuring and training, and the company assembles “tiger teams” for special projects involving AI. The company has also established Cloud Centers of Excellence to lead the adoption of new technologies, establish governance, and identify AI use cases. “We get a lot of input and support from the AWS team,” says Lystad. “They are present in our internal AWS communities when we have meetings or training webinars, and they also help facilitate workshops by contributing architectural advice.”

By focusing on all four pillars of its AI strategy, Visma has supported more than 600 AI initiatives and has upskilled 4,000 employees. As a result, the company has seen a major improvement in its operational efficiency, especially when it comes to accelerating coding-related tasks.

Roughly 62 percent of all Visma’s code benefits from AI-assisted review before being reviewed by engineers, leading to triple the number of production-ready deployments. “One of our team leads was so excited he was almost falling out of his chair because we are now innovating at a speed we’ve never seen before,” says Lystad. “We completed a feature set in 1 week that normally would have taken us 3 months.”

In turn, Visma’s engineers can spend more time on developing products that provide real value to its customers. Further, Visma’s leadership and business units have become more closely aligned in their brand and product goals. “Our vision now is to not create tools that you use for the job,” says Lystad. “We provide products that will do the job for you.”