Connecting AFAS to AI: leave, contracts and HR questions without a ticket
AFAS Profit can be connected to AI. IntraGPT reads data from HR, payroll and the rest of Profit, and can write back where that is allowed. Your administrator decides which parts take part, so the integration is exactly as broad or as narrow as you make it.
What the agent does with AFAS
Pull data out of Profit
The agent retrieves data from the parts released at your end, for example employees, contracts, leave balances or project hours. What is available is agreed when the connection is set up.
Filter instead of fetching everything
The agent searches on an employee number, a department or a period and keeps the answer manageable. Draining whole connectors does not happen: what was not asked for does not come out.
Write back into Profit
The connector can record a change or a request, for instance, where that has been released. That touches HR or financial data, so the agent only does it on explicit request and shows what it intends to send first.
Fetch documents
The agent retrieves a document attached to a record in AFAS, so it can summarise or quote an attachment without anyone having to download it first.
What organisations solve with the AFAS connector
HR
- Situation
- HR fields the same questions every day about leave balances, contract terms and collective agreement rules, and each one costs an exchange of emails.
- What the agent does
- The agent reads the balance and contract data from Profit and combines it with the policy documents in the knowledge base.
- Result
- Employees get an immediate answer and HR keeps its time for the cases that really need attention.
Payroll
- Situation
- Before the monthly run, changes have to be checked for deviations against the previous period.
- What the agent does
- The agent retrieves the period in question per department and lists the differences.
- Result
- The check starts with the outliers instead of with row one.
Project organisation
- Situation
- A project lead wants to know how many hours were booked on the project this month and by whom.
- What the agent does
- The agent queries the hours connector with a filter on project code and period.
- Result
- A current picture without the project lead needing access to Profit.
Management reporting
- Situation
- The management team wants a recurring overview of joiners and leavers per department.
- What the agent does
- The agent runs the same query every month and summarises the result in the chat or in a report.
- Result
- A steady rhythm without anyone having to build an export by hand.
How we set the connector up
-
1
Create a token in AFAS
An AFAS administrator creates an app connector with a token. That token travels as AfasToken in the header; IntraGPT stores it encrypted and never shows it back.
-
2
Agree the connectors
You decide which parts of Profit the agent may read and which it may update. Anything not on the connection does not exist as far as the agent is concerned.
-
3
Read only or also write
Most organisations start with reading. Writing back is switched on once it is clear which changes you want to be possible and who may request them.
-
4
Test on a copy
Before the connector goes into production we test the queries on the environment you use for that, so a filter that returns too much or too little surfaces earlier.
Employee data in an AI connector
AFAS holds employee and payroll data. That calls for a connector that returns no more than the question needs, and for a clear answer to the question of who may see what.
Where the data lives: Nederland of de EU
No more than the question requires
The agent filters on the fields the question needs and caps the number of records. There is no action that retrieves the whole workforce.
Permissions per role in IntraGPT
An agent for employees gets to see different data than an agent for HR. That way an employee can look up their own leave balance without reaching colleagues' data.
Every call logged
Calls and errors land in the connector log with the action and the outcome. For a change you can therefore trace when it was made and at whose request.
Own server, open models
IntraGPT runs on your own server with open-weight models. What comes out of AFAS does not leave that environment and is not used to train a model.
Questions about the AFAS connector
Can I connect AFAS to ChatGPT?
ChatGPT has no connection to AFAS, and employee data does not belong in a public chatbot. IntraGPT gives you the same way of working, but with a connector that may only call the agreed connectors, on your own server, with logging per call.
Does IntraGPT work with AFAS?
Yes. AFAS Profit is in the IntraGPT catalogue: read data, write back where that has been released, and retrieve documents out of Profit.
Do we have to configure connectors in AFAS ourselves?
Yes, and that is exactly its strength. Your own configuration determines which data comes out, so you stay in control of what the AI gets to see. If something is already released for reporting or another integration, it can often be reused.
Can the AI change anything in AFAS?
Only if you release writing for that, and only on explicit request. The agent first shows what it wants to write. If you release nothing to write, the integration is technically read only.
What does an AFAS connector cost?
The connector is part of the project we run together. The work sits mainly in aligning the connectors and the permissions, not in the technology. In a thirty minute call we look at the connectors you already have.
How long does setup take?
If the configuration already exists, the connector can be tested the same day. If something still has to be released in AFAS, the AFAS administrator's diary sets the lead time. Start with one subject that solves a lot of questions, such as leave or contract data.
Can an employee look up their own data without seeing anyone else's?
Yes, provided the configuration in AFAS allows it and the agent sits on the right role. That is a configuration choice in AFAS and IntraGPT together; we walk through it explicitly during setup.
Can we combine this with documents from SharePoint?
That is the most common combination. The data from AFAS answers the factual question, the documents in the knowledge base supply the rule or the policy. The agent always states which source came from where.