Visma AI Connect, Business Central Copilot or a shared data platform?

Choose by the boundary of the work

An ERP assistant is a good starting point for work inside its ERP. A shared data platform becomes relevant when the outcome needs several systems, reusable definitions and an application or workflow that keeps running. These approaches can coexist.

Visma AI Connect

Visma AI Connect connects AI clients such as Claude or Copilot to Visma Net through MCP and respects the user’s permissions. Visma describes supported reads and updates across several ERP areas, with one company per session. It is a practical route for interacting with that ERP.

Check Visma’s current documentation for object coverage and limits. An ERP tool connection does not by itself define a shared customer model across an unrelated CRM and accounting system.

Business Central Copilot

Microsoft’s Copilot capabilities are integrated into Business Central online. Available features depend on the capability, language and region, and Microsoft provides developer tools for extensions. Evaluate the specific task in the tenant and release you use.

A team standardized on Business Central should test its included capabilities first. Additional data integration and operating design may still be needed when the workflow extends beyond that application.

Where Ronja fits

Ronja provides a shared data foundation and a runtime for AI-built applications and workflows across existing sources. AI helps build the solution; Ronja operates it with access controls, reusable definitions and execution history.

For example, a receivables process might combine ERP invoices, CRM responsibility and a company-specific review step. The implementation still needs source access, a customer mapping, explicit calculations and an owner. Ronja does not make those business decisions disappear.

Where Fabric and Databricks fit

A company with an established data engineering team may already use Microsoft Fabric or Databricks as its data platform. Treat the decision as an architecture and ownership question: which layer stores and transforms data, where shared definitions live, and who operates the end-user solution?

The relevant comparison is the complete workflow you need to maintain. A product analogy is not evidence of feature parity. Ask each supplier to demonstrate your required connection, calculation, permissions and failure handling.

A practical evaluation

Start with one real task, one source dataset and a written acceptance criterion. Compare the time and ongoing responsibility required to reach a reconciled result, rather than the fluency of the first answer.

Verify where credentials and data are processed, which actions require approval, how changes are reviewed and what happens after a source schema changes. Record unsupported requirements explicitly before making the choice. Product coverage changes; the linked primary sources were reviewed on 29 September 2026.

Sources

• Visma: AI i affärssystemet
• Microsoft: Copilot in Business Central
• Microsoft: Copilot developer tools
• Ronja platform