AFAS groups its AI features under the name Jonas. The assistant summarizes workflows, translates replies, fills in workflow fields, and answers questions based on documents you point it to. All of that happens inside your AFAS environment. If you want AI to work with data from AFAS and other systems, you build your own application on the AFAS Profit connectors, with a person approving every change.
AFAS Software is a Dutch ERP vendor, and Profit is its suite for HR, payroll, finance, and workflows. AI use is growing fast. According to Eurostat, 20% of EU enterprises with 10 or more employees used AI in 2025, up from 13.5% a year earlier. In the Netherlands the figure was 33%, according to the Dutch statistics office CBS (in Dutch).
What Jonas can do in AFAS today
The AFAS Help Center (in Dutch) describes three ways Jonas works in Profit:
- Jonas automation analyzes the data in a workflow and fills in fields automatically. The workflow administrator decides which fields and with what.
- Jonas assistant runs a fixed instruction (a prompt) at a chosen point in a workflow. Users see the result in an information block, where they can read it, copy it, and discuss it.
- Jonas automation for advisory files analyzes a tax question in Profit Fiscaal, the tax module, using external legal sources.
Jonas also recognizes receipts for expense claims in AFAS Pocket, the mobile app, and helps write replies in InSite, the employee portal. When you compose a workflow reply, you can ask Jonas for a summary, a translation, or your own instruction. On its AI product page, AFAS also lists translation into languages including English, French, German, Spanish, Polish, and Papiamentu, matching vacancies with applicants, and reviewing purchase quotes through a workflow.
A press release from AFAS dated April 17, 2025 says more than 1,000 organizations were using Jonas at the time. One example is Zorggroep Sint Maarten, a Dutch care group. There, Jonas drafts answers to employee questions for the HR, payroll, and application management teams. The team approves the draft and sends it on. According to René Verhoef of the care group, the approach saves about six hours a month.
What Profit 8 added
According to the AFAS customer portal, Profit 8 was scheduled for delivery in mid-July 2026. For Jonas, that release brings these new features:
- Chat with Jonas inside a workflow, using source files as context. The chat is personal, so other workflow users don't see it.
- Fill in workflow fields automatically with AI, such as a summary, an end time, a project, or a priority.
- The Jonas assistant (previously called Jonas instruction) now reads PowerPoint, Excel, CSV, and JSON files and can use the Profit Help as a source.
- During imports, Jonas maps import fields to Profit fields automatically. The user checks the mapping.
- Jonas insight shows usage for all the new features.
Jonas is moving deeper into daily work in Profit. Which features you can use depends on your version, your package, and how your environment is set up.
Where Jonas reaches its limits
Jonas is built for work inside AFAS. That makes it strong in workflows and files, and it also sets the boundaries. The AFAS documentation points to five limits:
- Only the data you send. On its AI topic page, AFAS says customers decide which information goes to Jonas. AFAS doesn't describe a feature that lets Jonas pull data from your CRM, your scheduling system, or a production system.
- Limited context in replies. In a workflow reply, Jonas only processes the data visible on the page, without attachments. This feature works in InSite and OutSite. The Pocket app doesn't support it.
- Always a proposal. Jonas never adds a reply by itself. It shows a suggestion, and AFAS asks you to always check the text.
- Activation and setup. You activate Jonas per environment in the Profit properties. AFAS advises doing this outside working hours and in a test environment first. Some features need extra setup, and a fair use policy applies.
- No stored dialogue. Instructions and answers in workflow replies aren't stored in Profit. Token usage is.
AFAS is open about the technology. According to AFAS, Jonas runs on Microsoft Azure Foundry, in AFAS's own business tenant, within the European Economic Area. AFAS states that customer data isn't used to train the model. For speech recognition, AFAS hosts its own open source based models inside AFAS Online.
What you build when AI has to work across AFAS
Many processes don't stop at AFAS. Onboarding also touches scheduling and access to other systems. A purchasing process may run through a supplier portal or another ERP. If you want AI to support a process like that, you need software that combines data from several sources.
| Task | What Jonas does in AFAS | What you build yourself |
|---|---|---|
| Answer an employee question | Drafts an answer in the workflow | An assistant that also searches other systems and cites its sources |
| Fill in fields | Fills fields in a workflow | A check against data from other systems before anything enters AFAS |
| Analyze documents | Reads documents you point to in the workflow | A processing step for documents that arrive outside AFAS |
| Import data | Maps import fields to Profit fields | A permanent integration that exchanges data automatically, with checks |
| Apply changes | Makes a suggestion, the user decides | An approval screen, then a write through an UpdateConnector |
The technical foundation is the AFAS Profit API. An external application reads data through a GetConnector. Through an UpdateConnector, it adds, changes, or deletes records. Access runs through an App connector: there you generate a token and decide which connectors the application may call. According to the REST API documentation, you can only call connectors that are activated in the App connector. Data comes as JSON over REST or as XML over SOAP, in batches of 100 rows per request by default.
Our article on ERP system integration explains how to set up a connection like this. For the wider question of preparing your systems for AI, see opening up your application landscape with APIs for AI.
What your own AI application on AFAS looks like
An application that combines AFAS with other systems usually follows four steps:
- Read. The application uses GetConnectors to fetch exactly the fields it needs, plus data from the other systems.
- Propose. A language model organizes that data, summarizes it, or suggests a change, with a reference to the source.
- Review. A staff member sees the proposal next to the current data and approves, edits, or rejects it.
- Write and record. Only after approval does the application write to AFAS through an UpdateConnector. Every step is logged: what the model proposed, who decided, and what changed.
A few design choices make the difference. Give the application its own App connector with only the connectors it needs. Send the language model only the fields the task requires. And respect the API's fair use policy: fetch data in chunks with Skip and Take, and don't request the same large dataset every minute.
This matches how AFAS thinks about Jonas. There, too, the AI proposes and a person decides. Our article on AI workflows with human approval shows how to design that review step. Start small, with one process and a clear measure of success. Our approach is in how to set up an AI pilot.
Where Isatis helps
For setting up Profit and Jonas themselves, you go to AFAS or your implementation partner. Isatis builds what surrounds them: software integration between AFAS and your other systems, and custom data and AI work, such as an assistant that works across several sources with an approval step for every change.
We have 30+ years of experience, 30+ engineers in Nijmegen and Sarajevo, and 100+ completed projects, and we're ISO 9001 and ISO 27001 certified. If AFAS is one of the sources in a process you want to support with AI, book a conversation. We'll look at which data the application needs.
Frequently asked questions
What is AFAS Jonas?
Jonas is the collective name for the AI features in AFAS software. It works in Profit, InSite, OutSite, and AFAS Pocket. It summarizes, translates, fills in workflow fields, answers questions based on selected documents, and recognizes receipts for expense claims. In April 2025, AFAS reported that more than 1,000 organizations were using Jonas.
How do you turn on Jonas in AFAS Profit?
You activate Jonas per environment in the Profit properties, on the Activation tab, and accept the terms. AFAS advises doing this outside working hours and in a test environment first. Nobody can work in Profit during activation. Some features need extra setup afterward.
Can Jonas use data from other systems?
Jonas works with the data in your AFAS environment and with documents you point to in a workflow or assistant. It doesn't query systems outside AFAS on its own. For that, you build an application that talks to AFAS through GetConnectors and UpdateConnectors and reads other sources as well.
Where does the Jonas AI run?
According to AFAS, Jonas runs on Microsoft Azure Foundry, in AFAS's own business tenant within the European Economic Area. Customer data isn't used to train the model. Speech recognition runs on models that AFAS hosts itself inside AFAS Online.
When should you build something alongside Jonas?
When a process needs data from several systems, or when AI has to prepare changes in AFAS that a staff member approves. A permanent, automatic exchange between AFAS and other software is also a job for a dedicated integration. Within a single AFAS workflow, Jonas is often enough.
