--- title: "Somebody exports the ERP into a spreadsheet every morning." url: "https://referent.app/industries/retail-consumer-goods" description: "Referent connects to the ERP, POS and inventory systems behind your operation, including custom platforms with no connector. Stock, orders and margin answers from live data." --- Retail & Consumer Goods # Somebody exports the ERP into a spreadsheet every morning. Everyone knows it is absurd and nobody has a way to stop it, because the system holding stock and orders was never built to be queried by anything else. Referent connects to it directly and produces the same numbers without the export. [Book a demo](/demo)[How it connects](#how-it-works) 0104How it connects ## It connects where your interface already connects. It maps the queries, the actions and the endpoints behind the interface, then connects at that level. Nothing to rebuild, nothing to migrate. Your team ### Questions and work, in plain language Chat, Slack, Teams or Telegram. Ask for a number, a report, a draft or a status. No new tool to learn. Referent ### The layer that understands your systems Maps queries, actions and endpoints. Answers from live data with citations back to the source. Executes the work on top: pulls the numbers, builds the report, drafts the follow-up. Your software ### The systems holding stock, orders and margin A custom or heavily modified ERP, POS exports arriving as files, supplier portals with no integration and the inventory spreadsheet that has become the real source of truth. * Custom ERP * POS exports * Inventory spreadsheets * Supplier portals * Warehouse system ### Standard tools too The 50+ connectors everyone advertises are included. That part is commodity. The hard systems are the point. * Microsoft 365 * Google Drive * Slack * Jira * Salesforce This is not RPA. No screen reading and no bots clicking buttons. Referent connects at the data layer, so a UI change does not break it. 0204Use cases ## Four answers that currently cost somebody their morning. Same numbers, same format, produced from the system instead of by hand. 01Real stock nowReal stock, right now, by SKU and locationNot the figure from last night's export. The current position, read from the system that holds it, with the reference so anyone can check it.02Morning reportThe morning report, without the morningSales, stock, orders and exceptions compiled on schedule in the exact layout your team already reads. The person who built it by hand gets their first two hours back.03Orders at riskOrders at risk, identified before the customer noticesShortfalls, delayed replenishment and unfulfillable orders surfaced from live data, with the follow-up to the supplier or the customer drafted and ready.04Margin questionsMargin and supplier questions answered on the spotLanded cost, price changes and supplier performance, pulled from the systems that hold them rather than reconstructed in a spreadsheet each time somebody asks. Live sessionERP · Warehouse system You: Real stock on REF-2210, by location, right now. * read ERP · stock * read warehouse system · positions Referent 1,880 units, but 640 of them are in the store network rather than the warehouse, and 120 are already picked for tomorrow. Warehouse 1,240 Stores 640 Picked for tomorrow −120 Sellable now 1,760 ERP · Stock · REF-2210 You: How does that compare with the overnight export? * read last night's export · 23:50 Referent The export says 1,880 sellable. It was taken before the pick, so it is 120 units optimistic on anything you promise today. Inventory export · 23:50 0304Deployment ## Days, not months. There is nothing to migrate and nothing to rebuild, so the timeline is short by construction. 1. 01 ### Map the system We watch how your existing app talks to the system underneath and document the queries, actions and endpoints behind it. 2. 02 ### Connect at the data layer Referent connects directly at that level. No connector to build, no migration, no change to the software your team uses every day. 3. 03 ### Validate with your team The people who know the process check the answers against reality. Permissions are mapped to the roles you already have. 4. 04 ### Live Your team asks in plain language and gets answers from live data, with citations. Scope grows from there. Reference deployment ### A mid-market manufacturer, on a custom ERP. Mapped, connected and in production in under a week, on a system no AI product could reach. System Custom ERP Time to production One week Migration required None 0404Security ## Built to clear the security review, not to argue with it. Most AI proposals die in IT. These are the properties that get the conversation past that point. * 01 ### Single-tenant isolation Your deployment runs in its own environment. Nothing is shared with another customer. * 02 ### Zero-retention models Prompts and responses are not stored by the model providers and are never used for training. * 03 ### Your permissions, mirrored Referent respects the access rules your systems already enforce. Nobody sees what they could not see before. * 04 ### No data duplication Answers are fetched from the live system at query time. We do not build a second copy of your database. * 05 ### GDPR European data protection requirements are part of the deployment, not an afterthought. * [ Read the security detail](/security) ## See it running on a system like yours. [Book a demo](/demo)[Sign in](https://dash.referent.app/onboarding) ## Be the first to hear about Referent news.