# AI for Sage X3 Purchasing and Stock: Orders, Receipts and Lots > Source: https://netray.co/erp-ai/sage-x3-purchasing-stock-ai > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: AI for ERP ## Short answer AI for Sage X3 purchasing and stock reads purchase orders, receipts, supplier data, stock balances and lots through the X3 GraphQL or web service APIs. It reviews orders, tracks supplier delivery performance and answers stock questions, while approvals and postings stay in X3. ## Kind use-case ## Erp Family other ## Erp Products - Sage X3 - Sage X3 Purchasing - Sage X3 Stock Management ## Industries - Food and beverage - Chemicals - Distribution - Light manufacturing ## Region Global ## Persona Purchasing Manager or Supply Chain Manager on Sage X3 ## Eyebrow Sage X3 purchasing + stock AI ## Intro Sage X3 is widely used by mid-size food, chemical and distribution companies that need multi-site, multi-company purchasing and stock control with lot tracking. Purchasing runs through requisitions, orders, receipts and invoice matching, and stock is managed by site, location, lot and quality status. In daily use, the friction is information access. A buyer wants to know which orders are overdue by supplier, a stock controller wants all stock in quality hold for a product family, and a finance user wants receipts not yet invoiced. X3 can answer each of these, but through different screens, inquiries or custom reports. AI can sit above those screens. It takes a question in plain language, translates it into an API query, returns the records and shows how it got there. It can also scan recent orders for oddities and watch delivery history for patterns. It does not replace X3 workflow or its approval rules. Sage also has its own assistant features, and for simple inquiry needs they may be enough. This page describes where a separate, private layer is worth the effort. ## Pain Points ### Overdue orders tracked by hand **Detail:** Buyers export open purchase lines and chase manually. Supplier patterns, such as always late on one product family, are not obvious. ### Stock in the wrong status **Detail:** Lots sit in quality control or rejected status longer than intended, blocking production or sales without anyone noticing. ### Receipts not matched to invoices **Detail:** Goods received but not invoiced accumulate, creating accrual noise at period end. ### Customized inquiries and reports **Detail:** X3 sites often rely on custom screens and Crystal-style reports that few people can modify. ### Multi-site visibility gaps **Detail:** Stock and demand across sites and legal entities is hard to see together when purchasing decisions need it. ## Use Cases ### Overdue and at-risk order tracking **Description:** Lists purchase order lines past or near their expected date, grouped by supplier, with delivery history and suggested follow-up. **Erp Objects:** PORDER, PORDERQ, supplier records, PRECEIPT **Outcome:** A prioritized chase list that replaces a weekly export-and-sort routine. ### Supplier delivery performance **Description:** Computes on-time and in-full delivery per supplier and product using order and receipt dates and quantities. **Erp Objects:** PORDERQ, receipt lines, BPSUPPLIER **Outcome:** Evidence for supplier reviews and for adjusting planned lead times. ### Stock status monitoring **Description:** Finds lots in quality control, rejected or blocked status beyond a set age and summarizes value and affected products. **Erp Objects:** STOCK, STOLOT, ITMMASTER, stock status A, Q and R **Outcome:** Earlier release or disposition of held stock; less hidden working capital. ### Lot trace questions **Description:** Answers where a given supplier lot went, which orders and customers received it and what remains on hand. **Erp Objects:** STOLOT, STOJOU, sales deliveries, receipts **Outcome:** Faster response to a recall inquiry, with the final decision made by quality staff. ### Purchase order anomaly review **Description:** Flags lines whose price, quantity or unit differ from history or contract terms before approval. **Erp Objects:** PORDER, PORDERP, price lists, ITMMASTER **Outcome:** Fewer errors reaching the supplier and the invoice matching step. ### Goods received not invoiced analysis **Description:** Lists receipts without matching invoices by supplier and age, and suggests likely reasons such as price mismatch or missing invoice. **Erp Objects:** PRECEIPT, PINVOICE, accruals **Outcome:** Cleaner period-end accruals and less manual reconciliation. ### Cross-site stock visibility **Description:** Lets a planner ask for availability of an item across sites before ordering more. **Erp Objects:** STOCK, sites, ITMFACILIT, transfers **Outcome:** Reduces duplicate purchasing when stock exists elsewhere in the group. ## Architecture **Overview:** A service reads X3 through GraphQL or web services using a dedicated, scoped user, stores what it needs for analysis in your environment and serves a local model that explains results. Anything beyond reading is proposed and approved. **Layers:** #### API access **Detail:** X3 exposes GraphQL through the Syracuse server, along with SOAP web services published for the folder; both can be used for read access. #### Folder and site mapping **Detail:** X3 data is organized by folder and site; the mapping carries legal entity, site and language into every query. #### Analytics store **Detail:** Order, receipt and stock history held in a store you control for delivery performance and trend calculations. #### Model serving **Detail:** Open-weight model hosted on vLLM or Ollama on your hardware or sovereign cloud. #### Approval gate **Detail:** Any write, such as a changed expected date, is created as a proposal and applied by an authorized user through X3. ## Deployment Options ### On-prem X3 with on-prem AI **Fit:** Companies hosting X3 themselves **Detail:** Store and model sit in the same network as the Syracuse server. ### Sage-hosted or partner-hosted X3 with private AI **Fit:** Cloud X3 customers **Detail:** API access is arranged with the host, and processing stays in your tenancy. ### Regional deployments **Fit:** Groups with EU or other residency rules **Detail:** Regional nodes keep local entity data local; a restricted summary can feed group reporting. ## Compliance ### Lot trace queries read X3 data to support your trace procedures; the official records remain in X3. **Framework:** FSMA and EU food traceability ### Product and lot attributes in X3 can be queried; regulatory determinations stay with your compliance team. **Framework:** REACH and chemical regulations ### Approval limits and segregation of duties in X3 remain unchanged; AI does not approve or post. **Framework:** Financial controls ### Supplier contacts are excluded from prompts unless a use case requires them, and data stays in your chosen region. **Framework:** GDPR ## Integration Notes - Check your X3 version: GraphQL availability and web service options depend on the release and the Syracuse setup. - Create a dedicated user with limited function profiles; queries should inherit the asking user's rights wherever possible. - Specify the folder in every call; multi-folder environments will otherwise return the wrong data. - Table names such as PORDER and STOCK are stable, but customer-added fields (such as those with the X prefix) need to be mapped. - Use incremental extracts on update dates for history; avoid heavy reads in the posting windows used by finance. - Where X3 is integrated with a WMS or EDI, include those event sources so receipts and shipments look complete. ## Product Fit ### ERPray **Why:** Plain-language purchasing and stock questions with the query shown and X3 access respected. ### Custom build **Why:** Supplier delivery models, stock status monitors and anomaly rules tuned to your items and contracts. ## Engagement ### Discovery **Duration:** 1 week **Deliverables:** - X3 version, folder and hosting review - Top buyer and stock controller questions - Access and security plan ### Connect and map **Duration:** 2 weeks **Deliverables:** - Scoped API user - Mapping for purchasing and stock objects - Custom field register ### Pilot **Duration:** 3 to 5 weeks **Deliverables:** - Overdue order tracker - Stock status monitor - Accuracy check against X3 inquiries ### Extend **Duration:** Ongoing **Deliverables:** - Supplier delivery scorecards - PO anomaly review - Monitoring and audit log review ## Buyer Questions - Which X3 version and API will you use, and is it supported on our hosting? - Do queries run with the asking user's X3 rights? - How do you handle multiple folders, sites and legal entities? - How are custom fields and screens included? - Can I see the query behind every figure? - Is any data sent to a public AI service? - How are changes proposed, approved and logged? ## Faq ### Does Sage X3 have built-in AI? **Answer:** Sage has introduced AI assistant features across its products. Check what your X3 version and subscription include. A separate private layer is useful when you want models on your own infrastructure, cross-site analytics or custom checks such as purchase order anomaly review. ### Which X3 interfaces can AI use? **Answer:** Typically GraphQL through Syracuse and published SOAP web services for reading, and database replicas in some self-hosted cases. Availability depends on your version and hosting, so we confirm during discovery before promising any approach. ### Will AI approve or post purchase documents? **Answer:** No. Approval and posting stay with authorized users in X3. AI can flag issues and draft suggestions, but the transaction is made by a person, preserving your control and audit evidence. ### Can it handle several legal entities? **Answer:** Yes, if the mapping carries folder, company and site into every query and respects each user's access. Cross-entity views need explicit authorization, and we test for data leakage between entities. ### How does it help with lot tracking? **Answer:** It answers questions about where a lot went and what remains, using stock journal and receipt data. It speeds up the lookup. The formal recall decision and records stay with your quality team. ### What is the effort for a small team? **Answer:** A read-only pilot for purchasing and stock typically takes 6 to 8 weeks. Heavier customization, several folders or restricted hosting access extends that. ### Is a GPU needed? **Answer:** For a handful of users, one GPU server with a mid-size open-weight model is usually enough. Sizing follows the number of concurrent users and the amount of document retrieval you want. ## Search terms sage x3 purchasing ai, sage x3 stock ai, sage x3 inventory ai, sage x3 purchase order automation, sage x3 supplier performance, sage x3 graphql, sage x3 lot traceability, sage x3 overdue purchase orders, ai for sage x3, sage x3 private ai, sage x3 stock status quality, sage x3 goods received not invoiced --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json