AI inside AFAS: how an AI coworker natively integrates with Profit

AFAS is different

AFAS Profit is the dominant ERP in the Dutch mid-market. For financial administration, order management, project admin and HR processes it's often the only system. That makes the integration critical, what happens in AFAS happens at the heart of the organization.

Many automation tools work around AFAS: pull data out of AFAS, process it in a separate tool, push it back in via CSV or by hand. That works, but it doesn't scale. Every change to your AFAS configuration forces a change to the connector. Every new business rule is maintained in two places. And the audit trail ends up fragmented.

What native integration means

Our AI coworkers talk directly to AFAS via GetConnectors and UpdateConnectors. No intermediate files. No separate scheduling. What the AI coworker sees and does is the same as what a human sees and does, including permissions, fields and validations.

In practice that means:

• Postings, sales orders and project entries land straight in Profit, with the correct audit trail at the user level.
• Master data (customers, items, contracts) is pulled in real time, not from a copied snapshot.
• Changes to your AFAS configuration, new fields, new validations, new workflows, come along automatically.
• Authorizations follow the AFAS roles you already have. The AI coworker gets its own user, with explicit rights.

Processes we frequently build in AFAS

Invoice processing with three-way matching against purchase orders and receipts. Order processing with validation against customer pricing agreements and stock. Project entries, hours, costs, revenue, based on emails and PDFs from subcontractors. Service-ticket processing linked into Bouwworks or AFAS project admin itself.

In all of these the same pattern applies: the AI coworker works within the boundaries you've already defined in AFAS. We don't add a new rules layer; we make the existing process executable without manual intermediate steps.

What stays outside AFAS

AFAS is strong on data and administration. For intake (email with attachments, portals, scanned documents) and classification, our AI coworker runs on our own platform. That's where extraction, validation and escalation handling happen. Once the outcome is settled, the AI coworker writes through to AFAS.

Your team watches along on our workflow and case dashboards, where every case shows up with document, extracted data, validation result and proposed AFAS action. Approvals and exceptions flow through that dashboard; the execution happens in Profit.

What you get out of it

For a typical invoice-processing flow in AFAS: posting time per invoice from minutes to seconds, matching rate from below 70 percent to above 90 percent, and auditability down to the case level. No export, no middleware, no divergent audit trail. Just AFAS, minus the manual intermediate steps.