Data to AI™ / ChatGPT/AFAS
Check whether your AFAS figures are fresh and complete before you trust them, then ask ChatGPT the finance question, approving each lookup as it runs.
Better answers, fewer tokens.
ChatGPT
Which article groups had the most articles modified last week, and what is their average sales price?
ArtGroup 100 changed most, with 86 articles by ModifiedDate and an average SalesPrice of €42.10. All three groups shown were edited during last week's price update.
| ArtGroup | Articles by ModifiedDate | Average SalesPrice |
|---|---|---|
| 100 Profit_Article | 86 | €42.10 |
| 200 Profit_Article | 54 | €118.40 |
| 300 Profit_Article | 21 | €9.75 |
AFAS has no date range, so each run pulls current articles up to the Limit; the week is counted on ModifiedDate, and Blocked articles are left out of the average.
AFAS Profit is Dutch business software for finance, HR, payroll and CRM. Dataddo extracts whichever GetConnector you have set up in AFAS, with the fields you pick, custom fields included. The supplied GetConnectors shown here, such as Profit_Article and Profit_Employee_Leaves, are examples from the AFAS documentation.
Run data_status on the AFAS flow and tell me the latest dataddo_extraction_timestamp. Did yesterday's scheduled run happen?
Before sending figures to the board, the controller confirms each tool call and sees whether the AFAS data is from this morning or from last week.
How many rows did the last AFAS extraction return? If the count equals the Limit of the source, warn me that records may be missing.
A financial analyst catches a run that silently stopped at the Limit and asks the Dataddo admin to raise it before the numbers are used.
Which ArtGroup in Profit_Article has the most Blocked articles? Show the top ten groups with their ItemCode counts.
The controller spots groups where many articles are blocked but still listed and raises it with purchasing in the monthly review.
Data to AI™ is priced per seat, like the AI assistants your team already uses, with ~25% off on annual billing. Need tailored contracts or volume pricing? Talk to us about Enterprise.
Serve governed business data to Claude, ChatGPT and other AI tools over MCP.
$10.49/month
per seat
3 seats minimum
Flexibly deploy Dataddo in any cloud or hybrid environment.
Custom
We have a payment model that works for you
| Without Dataddo |
With Dataddo
|
Outcome for you | |
|---|---|---|---|
| Trust in the figures | An uploaded spreadsheet gives ChatGPT no hint of when the AFAS data was exported or whether rows were cut off. | ChatGPT calls data_status for the last extraction and reads dataddo_extraction_timestamp on every row, so the answer states its age. | Data age stated in every answer |
| Personal data from HR and payroll | A full GetConnector export lands in the conversation with every column AFAS returns, names and addresses included. | You choose the fields in the Attribute picker when you create the source, for example only StartDate, Year and EmployeeId from Profit_Employee_Leaves, so other columns never reach ChatGPT. | Only chosen columns shared |
| Control over each query | Nobody can say which part of an attached file shaped the summary ChatGPT wrote. | ChatGPT waits for your confirmation before each Dataddo tool call, and individual tools can be switched off in the app settings. | Every lookup approved |
| Token consumption | ChatGPT spends tokens exploring raw AFAS columns, sampling rows and retrying until it understands the data. | Field definitions and relations are supplied up front, so the model skips the exploration and goes straight to the answer. | Fewer tokens per answer |
| Answer quality | Prone to wrong joins, invented fields and numbers that are hard to check. | Answers grounded in defined fields. In the Dataddo benchmark, 78.5% of questions were answered correctly, against 65.5% with plain CSV files. | Better answers |
The AFAS datasets finance and RevOps teams use most, with their real field names. ChatGPT queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask ChatGPT
|
|---|---|---|
| Profit_Article (example GetConnector) |
|
Which ArtGroup holds the most Blocked articles? |
| Profit_Employee_Leaves (example GetConnector) |
|
Which EmployeeId has the most leave records this Year? |
| ProfitGetConnectoren (example GetConnector) |
|
Which GetConnectors are Geblokkeerd, so they cannot be used for extraction? |
| Date and key fields |
|
Which records have a last modified on date of yesterday in our AFAS data? |
We asked 43 pre-registered questions of three real datasets - Google Search Console, HubSpot deals and Google Ads - and delivered the same rows to the same model in four ways. Only the data layer changed.
of questions answered correctly - the best of the four delivery methods tested
fewer confidently wrong answers - 14.0% of runs against 24.0% and 26.3%
more correct answers on messy CRM data - better on 8 questions, worse on none
Same model, prompts and rows in every condition; 759 graded runs, ground truth frozen before any run. On Google Search Console alone, documented CSV files did as well as Data to AI™ - the advantage comes from messier data such as HubSpot deals.
Read the benchmark →Create an AFAS source in Dataddo with your AFAS token, Environment ID and a GetConnector, and attach the flow to a ChatGPT destination or AI Model. In ChatGPT on the web, enable Developer mode in Settings > Security and login, create a Dataddo Data Access app with https://headless.dataddo.com/mcp-data and OAuth.
The ones you have set up in AFAS and extract in Dataddo, including your own copies and custom fields. Profit_Article and Profit_Employee_Leaves on this page are only examples of supplied GetConnectors.
Developer mode runs in ChatGPT on the web on Pro, Plus, Business, Enterprise and Education. On Business and Enterprise, a workspace admin may need to allow it first.
Yes. ChatGPT asks you to confirm each tool call, so you approve every lookup in your AFAS flows before any result appears in the chat.
As current as the last run. AFAS has no date range, so each scheduled run pulls the records available at that moment, and data_status tells ChatGPT when the flow last ran.
Yes, with a JSON filter as described in the AFAS documentation. The filter can only be set when you create a new source, so set it up after your historical load and delete the original source.
Pick only the fields you need in the Attribute list when you create the AFAS source. ChatGPT can query nothing beyond what reaches the flow, and Dataddo writes nothing back to AFAS.
ChatGPT can call known_gaps, which lists what the data cannot answer. If the GetConnector you extract has no modification date such as ModifiedDate, for example, questions about recent changes are out of reach.
