Yes, if the basics are clean: real supplier lead times, open purchase orders and stock levels you trust. Your ERP's reorder rules do the arithmetic. AI adds what the rules cannot see, such as supplier delay emails and large customer orders mentioned in messages, explains each warning in plain words and drafts the reorder. A planner approves every order.
#Key takeaways
- Most stock-outs trace back to stale lead times, stale reorder points and late supplier news, not to a weak forecast model.
- ERP reorder rules and statistical forecasting already exist; check you are using them before adding AI.
- AI is useful for unstructured signals: supplier delay notices, customer emails and sales notes.
- Every reorder suggestion should come with its reason, and a planner approves it.
- Measure stock-outs on your most important items and how often planners accept or change suggestions.
#What does a planner's week look like before a stock-out?
On Monday the planner exports a stock report from SAP, Oracle or Odoo into Excel and compares it with min and max levels set long ago. A supplier emails that a shipment slipped; the email sits in a shared inbox. A sales manager mentions a large order on WhatsApp. The open purchase order still shows the original delivery date, so nothing looks wrong.
Later the shelf is empty, and everyone finds the warning in their inbox. The data was there, just not where the reorder logic looks. Retail research points the same way: IHL Group, september 2025 names supplier missteps and internal inefficiencies among the main causes of out-of-stocks and overstocks.
#What has to be right before AI can help?
- Lead times measured from actual receipts, per supplier and item, not the number in the supplier master.
- Open purchase orders with current confirmed delivery dates.
- Stock levels you trust, including stock in transit and in other warehouses.
- Items classed by importance, so the most critical lines get the closest watch.
- Reorder points and safety stock reviewed recently, not inherited.
- A named planner who owns each item group.
Skip this and AI only produces confident warnings on bad data. Gartner, july 2024 named poor data quality among the main reasons generative AI projects are abandoned after proof of concept.
#What does AI add that reorder rules do not?
Reorder rules work on numbers in the ERP. AI can read what is not there yet: supplier emails announcing delays or partial shipments, customer emails and sales notes about unusually large orders, and promotion plans in a shared document. It turns each into a proposed change, such as a new expected date on an open order, with a link to the source message.
It can then explain a warning in plain words: this item will run out before the next receipt because the supplier moved the shipment and a large customer order is due. And it drafts the reorder, or the expedite request to the supplier, for the planner to approve, edit or reject. Nothing is ordered without the planner.
#When are your ERP's reorder rules enough?
If supplier news reaches the ERP reliably, lead times are measured and the item range is stable, your ERP's replenishment rules or MRP, such as Odoo replenishment or SAP MRP, are enough. A dedicated demand planning tool fits when the forecasting itself is the problem, for example strong seasonality across many locations.
AI fits the gap between the inbox and the ERP: when the warnings exist but arrive as messages in Arabic and English, across many suppliers, and nobody has time to read them all against open orders.
#How does Stennir run this?
Stennir starts with AI consulting, because the honest answer is often to fix lead times and reorder points first. The AI Opportunity Audit ranks five workflows, each with a cost and a payback case, and recommends one build, sometimes a data fix rather than an AI one.
When the inbox-to-ERP gap is the problem, AI implementation builds the reading of supplier and customer messages, the proposed updates and the drafted reorders, with a planner approving each. Arabic and English are fully supported, other languages on request. Managed AI Operations then tracks how often planners accept suggestions and fixes what they keep rejecting. AI training helps planners work with it.
We are
Stennir connects supplier and customer messages to your reorder process, with a planner approving every order and every change to an expected date.
We aren't
Stennir is not a demand planning suite and does not place purchase orders on its own.
#What should a supply chain manager do next?
List your last five stock-outs on important items and find where the first warning appeared. If it was in an email or a message, the gap is reading, not forecasting. See AI for operations and how to pick the first process to automate, then take the free readiness assessment or book a 30-minute call.
#Common questions
What do you need from our team?
Read access to stock, open orders and receipts in your ERP, the shared inbox where supplier news arrives, and the planners who own item groups. They walk through recent stock-outs, so the build targets the signals that were actually missed. A named planner approves every suggestion.
Will it place orders automatically?
No. It drafts reorders, expedite requests and proposed date changes on open orders. A planner approves, edits or rejects each one, and the decision is logged with its reason. Some teams later allow automatic updates for low-risk changes, but that is your decision, made with the evidence in hand.
Does it read supplier emails in Arabic?
Yes. Stennir builds workflows that read Arabic and English messages, including mixed ones, and link each to the matching purchase order. Other languages are supported on request and tested on your own supplier messages first. Unclear messages go to a planner rather than updating anything.
What could go wrong?
Wrong lead times produce wrong warnings, and a misread email can suggest the wrong date. That is why data comes first, every suggestion cites its source and a planner approves it. Track rejected suggestions; a rising rate means the inputs need attention before anyone trusts the output.
