# How to use ChatGPT with QAD Adaptive ERP

> Source: https://netray.co/integrations/qad-chatgpt
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Integrations

## Short answer

You can use ChatGPT with QAD in three ways: paste or upload exports, give a custom GPT an action that calls a small API in front of QAD, or connect through an MCP-style connector. For regulated or export-controlled data, many QAD sites choose a private model instead.

## System

QAD Adaptive ERP

## Capability

chatgpt

## Intro

People on QAD ask the same question: can I just ask ChatGPT about my orders and inventory? The answer depends on how the data reaches the model, and that is a data governance question before it is a technical one.

ChatGPT by itself knows nothing about your QAD instance. It becomes useful when it can see current data, either because you hand it an export or because you give it a controlled way to call your systems. The right choice depends on the sensitivity of the data and how live the answers need to be.

This page covers the realistic options for QAD Adaptive ERP, from the ten-minute export to a governed API, and when to stop and use a private model.

## Use Cases

### Explain a messy export

**Description:** Upload an MRP action message report or an open PO listing and ask for the five lines that most need attention this week, with reasons.

### Draft supplier communication

**Description:** Give ChatGPT a late-delivery list and your tone guidelines and have it write one email per supplier for a buyer to review.

### Translate QAD jargon for new staff

**Description:** A planner can ask what a field or screen means in plain language. Use your own training notes as supporting material rather than relying on the model's general knowledge.

### Analyse a month of scrap

**Description:** Upload a scrap transaction extract and ask for patterns by item, work center and shift, then ask it to propose follow-up questions.

### Customer order status via a custom GPT

**Description:** A custom GPT with an action calls a read-only endpoint you host, so a customer service rep can ask for a status and get live answers.

### Prototype a new report

**Description:** Describe the report you want and let ChatGPT draft the SQL or the Power BI measure, then run it against a test database yourself.

## System Surface

- Browse and export: QAD browses and reports can export to CSV or Excel, the simplest path for ad-hoc analysis.
- OpenEdge SQL: JDBC or ODBC read access to a replica lets you build an API that returns only what the GPT needs.
- QXtend and QAD web services: callable endpoints for specific business data and transactions, subject to your release and licences.
- Your own wrapper API: a thin REST service with an OpenAPI description is what a custom GPT action actually calls.
- Scheduled extracts: nightly files to a secure location give a snapshot model that avoids live connectivity.

## How It Works

### Classify the data

**Detail:** Decide which QAD data may leave your network at all. Customer pricing, defense program data and export-controlled drawings often cannot.

### Start with exports for low-risk analysis

**Detail:** Use sanitised extracts with identifiers removed. Check your organisation's ChatGPT plan and data controls first.

### Build a wrapper API for live access

**Detail:** Expose a few read-only endpoints such as item availability and order status, each backed by a parameterised query against a replica.

### Describe it with OpenAPI

**Detail:** Write clear operation descriptions. A custom GPT chooses actions largely from those descriptions.

### Authenticate properly

**Detail:** Use OAuth or per-user tokens so each call is tied to a user and QAD-level permissions can be applied in your wrapper.

### Test with real questions

**Detail:** Collect fifty questions your staff actually ask and check answers against QAD screens before rollout.

## Example Flows

### Customer service order lookup

**Trigger:** A rep asks the custom GPT where order number 1234 stands.

**Steps:**

- The GPT calls the order status action with the order number.
- Your API queries the replica for lines, ship dates and open quantities.
- The API returns a compact JSON object without cost or margin fields.
- The GPT writes a short status the rep can read to the customer.

**Result:** The rep gets an accurate answer in seconds and cost data never leaves your network.

### Weekly scrap review from an export

**Trigger:** A quality manager exports scrap transactions on Monday.

**Steps:**

- The file is stripped of operator names and customer references.
- It is uploaded to an approved ChatGPT workspace.
- The manager asks for top contributors and repeated defects by work center.
- The summary is pasted into the weekly quality meeting notes.

**Result:** A review that took an hour of spreadsheet work is a draft in minutes, with a person verifying the numbers.

## On Prem

ChatGPT is a hosted service, so by definition the content you send leaves your network. For many QAD users that is acceptable for sanitised data and not acceptable for controlled technical data, customer-confidential pricing or contracts that restrict where data may be processed. If your QAD runs on-prem for exactly that reason, the same logic usually applies to the assistant. The alternative is a private model served locally with the same wrapper API pattern: the user experience is similar, the model is open-weight, and the data stays within your boundary. Some teams use both, ChatGPT for generic drafting and a private model for anything that touches QAD records.

## Security

- Read your plan's terms on training use, retention and admin controls before any QAD data goes in.
- Expose only read-only endpoints to a custom GPT and never give it credentials broader than necessary.
- Strip or tokenise customer names, part numbers under export restriction and pricing from anything uploaded.
- Authenticate users individually so your wrapper can enforce QAD domain and site restrictions.
- Keep an access log for the wrapper API and review it periodically.

## Options

### Manual exports into ChatGPT

**Fit:** Fine for low-risk analysis and drafting with sanitised data.

**Effort:** Hours

### Custom GPT with a wrapper API

**Fit:** For live read-only questions when your data policy allows hosted processing.

**Effort:** 2 to 6 weeks including security review

### Private model with Netray or in-house

**Fit:** For controlled or customer-restricted data where nothing may leave the network.

**Effort:** Typically a scoped pilot in a few weeks

## Pitfalls

- Pasting production extracts without checking the data policy.
- Trusting arithmetic on large uploads. Verify totals independently.
- Giving the GPT an over-broad endpoint such as a generic SQL runner.
- Letting answers drift from QAD because the export was stale.
- Assuming consumer ChatGPT and the enterprise offering have the same data handling.

## Faq

### Is there an official QAD ChatGPT connector?

**Answer:** Do not assume one exists. QAD has been adding its own AI features, so ask your account team what is available for your release. Otherwise a wrapper API or file export is the practical route.

### Can ChatGPT query the QAD database directly?

**Answer:** Not by itself. It needs an action or connector that calls something you control. Giving it direct database credentials is a bad idea; use a read-only API with parameterised queries.

### Is it safe to upload QAD exports?

**Answer:** That depends on the data and your plan's terms. Sanitised operational data may be fine; controlled technical data and confidential pricing usually are not. Ask your security team first.

### Why would I use a private model instead?

**Answer:** Control. A locally hosted model keeps QAD data inside your network, which matters for defense, aerospace and customer contracts that restrict processing locations.

### Can ChatGPT update QAD records?

**Answer:** Technically a custom action could call a write endpoint, but we advise against it. Use drafted proposals and an approval flow that submits through QXtend with a named user.

## Search terms

qad chatgpt, chatgpt for qad, qad chatgpt integration, use chatgpt with qad erp, qad custom gpt, chatgpt qad adaptive erp, qad openai integration, chatgpt manufacturing erp data, qad erp chatgpt api

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json