The data foundation behind Addvant's AI ambitions

Addvant is a Stavanger-based accounting firm that manages bookkeeping, payroll, reporting, and advisory for 500 clients across Norway. Lately their attention has turned to AI, and what it can really do for an accounting firm.

Addvant's clients are spread across three different accounting systems: PowerOffice GO, Tripletex, and Visma Business NXT.

Working through the vendors' APIs one at a time has real limits. You get the data in the shape the vendor decided on, you can only reach it when and how they allow, and validating it properly is difficult. For an accounting firm that wants to build AI on top of that data, those limits are the problem.

We built Addvant a data layer of their own: one common data model in Azure that pulls accounting data out of all three systems and lands it in the same shape.

Addvant can now query the data freely, whenever they want, without asking three vendors for permission, and validate it in ways an API never allows. Because every client lands in the same shape, a tool only has to be built once: the same tool works for a client in PowerOffice GO, Tripletex, or Visma Business NXT. The AI runs on infrastructure Addvant owns and controls.

With the data layer in place, building tools on top gets a lot easier.

The first one is Addvant's own MCP server, which lets the accountants ask questions straight to the data.

With an MCP available, accounting data can be used in a much broader way than before. It can be analyzed to understand the financial situation, identify trends and developments, and uncover issues that may need attention.

At the same time, accounting data can be combined with other sources, such as the company register, market data, LinkedIn campaign data and other relevant information. This makes it possible to put accounting data into context with the rest of the business and use it in processes where the data was previously not available, for example in sales, customer follow-up and analysis.