AI in the warehouse: what NG Copilot and MCP change in daily work
Most time in a warehouse is not spent carrying goods but looking for information. How an AI assistant that knows your orders, locations and stock history shortens that path — while people keep the decisions.
In short
- AI pays off most in a warehouse where people lose time today: looking for information about goods, orders and blockers.
- NG Copilot answers from NG WMS data and names the document behind every answer — wave, order, location or lot.
- Copilot proposes replenishment, a priority or a check, and the task is created only after a user with the right confirms it.
- The MCP server lets the AI assistants your team already uses work with NG WMS data, under the same permissions.
- AI does not replace a good warehouse model and scanning discipline — it builds on them.
Short answer: AI in the warehouse does not move pallets, but it shortens the path to information. NG Copilot is the AI assistant built into NG WMS: it answers questions about stock, orders, waves and lots in your own words and names the document behind every answer. It points out orders that are running late and proposes the next step, and a task is created only when a person confirms it. Through the MCP server, the AI assistants your team already uses work with the same data.
Where a warehouse really loses time
In most warehouses the most expensive minutes are not the ones on the forklift. They are the minutes when a shift lead is working out why an order is not ready, sales calls to ask whether the goods have arrived, and someone before a count leafs through item cards to understand where a variance came from. The answer is in the system, but it is spread across reports, documents and history.
That is where an AI assistant adds the most value: not by “smartly” running the warehouse instead of people, but by giving people in seconds an answer they would otherwise look for across ten screens.
Five questions asked every day
| Question | What Copilot does |
|---|---|
| What is late for dispatch today, and why? | picks out orders waiting for stock or carrying an ERP stop, with the wave and the reason |
| Where is lot L-0915 and who was it shipped to? | follows the lot from receipt to delivery notes and stock on a location |
| Which goods expire in the next 30 days? | gives a list by location and expiry, with the receipt documents |
| Which pick locations will run empty in the next wave? | compares reservations with stock and proposes replenishment from reserve |
| Where does the variance on location B-03-02-2 come from? | gathers receipts, issues and transfers for that location |
Questions are asked in your own words, in Serbian or English. The answer is short, and the documents behind it are one click away — so the answer is checked, not taken on trust.
A proposal, not an action on its own
A warehouse is a place where every stock change needs an owner. That is why NG Copilot does not change stock by itself. When it proposes replenishing a pick location, changing a priority or recounting, the proposal waits for confirmation by a user who holds the right for that action. Only then is a task created, and the history records who confirmed it.
The same applies to access: Copilot sees only the warehouses, goods owners and documents the signed-in user may see. A sales rep for one goods owner gets no answers about another owner’s stock.
What MCP is for
MCP (Model Context Protocol) is an open standard for connecting AI assistants and agents to business systems. NG WMS has an MCP server, so the AI tool your team already uses can read stock, orders and goods history without a separate integration for each tool.
The practical consequence: when your team changes AI assistant, the link to the warehouse stays the same — the same data, the same permissions, the same trail. MCP access is enabled per user, with a key the administrator can revoke at any time.
What AI cannot make up for
- A poor warehouse model. If locations have no purpose or capacity, AI cannot tell where goods belong either.
- Skipped scans. An answer is only as accurate as the record of how goods moved.
- Unclear rules. Minimum and maximum, the issuing rule and blocks are set by the warehouse; Copilot uses them, it does not invent them.
That is why AI in NG WMS stands on the same foundation as the rest of the system: the warehouse model, stock and replenishment and outbound with scan control. How Copilot and MCP look in practice is described on the NG Copilot and MCP page.
Frequently asked questions
What is NG Copilot in NG WMS?
The AI assistant built into NG WMS. It answers questions about stock, orders, waves and lots with the document behind the answer, flags dispatch blockers and proposes replenishment and priorities.
Can AI change stock levels on its own?
No. Copilot proposes; tasks, corrections and approvals are created when a user with the right confirms them. Every confirmation stays in the history.
What is MCP?
The Model Context Protocol, an open standard for connecting AI assistants to business systems. Through the NG WMS MCP server, your team’s AI tool works with warehouse data under the permissions of the user working with it.
Does Copilot see data for every goods owner?
Only what the signed-in user may see in NG WMS. The same rules apply to every MCP client.
Want to see how this would look in your warehouse?
In a demo we walk through your receiving, picking and dispatch in NG WMS and show how NG Copilot supports your warehouse team.