Netstock’s AI uses your ERP data to recommend inventory and demand planning actions. It never pulls information from the internet or guesses based on prompts. Before the AI processes anything, Netstock cleans and checks your data, recalculates risk for each SKU at every location daily, and highlights the decisions that matter most financially. Your team always reviews and approves every order before it goes into your ERP.
Key takeaways
- Built on your data. Recommendations come from your ERP’s sales, stock, orders, suppliers and lead times, synced nightly.
- Trained on SMB data. The AI learns from 42TB of proprietary SMB inventory planning data built over 15+ years.
- Cleaned before use. Netstock checks data quality and corrects history for distortions such as stockouts before forecasting.
- Recalculated daily. Forecasts, safety stock, supply risk and stocking policies update every day for every SKU and location.
- Prioritized for you. The AI flags stockouts, excess and surplus orders before you go looking, each with a suggested action.
- You stay in control. You set the policies, override recommendations and approve every order before it reaches your ERP.
- Kept in your own instance. Each customer has their own app and database, and Netstock runs its AI on its own infrastructure.
- Pooled learning is anonymous. De-identified data from many customers helps train the models and never identifies your business.
What data does Netstock’s AI use?
Netstock’s AI uses the inventory data already in your ERP. It works from a daily copy of that data, synced automatically every night through the broadest range of over 60 ERP integrations in the market. Your ERP stays your system of record.
The data that drives planning includes:
- Products and locations: your product master, warehouses and branches
- Stock: current stock on hand by location
- Sales history: sales and issues by month
- Open orders: outstanding purchase, sales, production and transfer orders
- Suppliers and lead times: active suppliers, plus completed purchase orders that show how long deliveries actually take
- Bills of materials and supersessions: how finished goods link to components, and which products replace others
The AI doesn’t use public internet data, and it doesn’t rely on whatever someone happens to type into a chat box. Every recommendation traces back to your own numbers.
How does Netstock make ERP data fit for AI decisions?
Raw ERP data is very seldom ready to be used by AI, since most ERPs have accumulated a number of oddities over the years, such as zero cost prices, negative stock levels, missing lead times, and orders which were never actually closed. If an AI is given that kind of data, it will produce answers with confidence based on poor inputs. Netstock transforms ERP data into data suitable for making decisions in three steps.
- It detects poor-quality data. Netstock carries out a wide range of specific validation checks on your ERP data and rates the quality of that data. At the same time, it highlights the problems that have the greatest effect on your results, for example cost prices of zero or less, negative stock on hand, items with no lead time or no supplier, and overdue orders.
- It overlooks anomalies which would distort the plan. Automated rules eliminate data that does not match reality. For instance, a purchase order remaining open in the ERP for nine years will not be included as stock during the process.
- Showing true demand involves correcting history. If a product was out of stock, the sales history will not accurately represent customer demand. By taking account of the stockout periods, Netstock adjusts the historical sales figures so that the forecasts show the actual demand rather than just the amounts that were available for sale.
From there, Netstock recalculates forecasts, safety stock, supply and forecast risk, and inventory stocking policies every day for every SKU at every location. These calculations are stored in Netstock, not in your ERP, so they stay current without adding load to your ERP. This is what lets Netstock flag excess stock and impending stockouts up to three weeks before they happen.
Data quality is something that everyone is responsible for. Your team is in charge of making sure that the data in your ERP is accurate, and Netstock tells you precisely which problems to address first.
How does the AI turn your data into recommendations?
Netstock’s AI works through your inventory data so your team doesn’t have to. It doesn’t wait for a question. Each day the AI looks across every item and location, finds the problems and opportunities, and suggests what to do about them, with the decisions that matter most to your bottom line first.
AI Opportunities flags nine types of issues automatically, each with a suggested action:
- Potential and actual stock-outs, with the fix: expedite an existing order, move excess from another location, or place an emergency order
- Surplus orders, with the fix: redirect the excess to another location, or cancel, reduce or delay the order
- Excess stock that could be redistributed to a location that’s running short
The AI Pack then explains what it finds, from executive summaries and item-level fixes to safety stock breakdowns, data troubleshooting and supplier email drafts. See the AI Pack features.
The AI also learns from how planners actually work. Netstock uses behavioral data from more than 2,500 customers to anticipate how users typically adjust recommended orders, and your own thumbs-up and thumbs-down feedback makes AI Opportunities more relevant to your business over time.
What is Netstock’s AI trained on?
Netstock’s AI is trained on the industry’s largest proprietary SMB supply chain dataset: 42 TB of data built from more than 15 years of real inventory planning. It reflects real ordering decisions, demand patterns and supplier performance from businesses like yours, not internet data or theoretical models.
That gives the AI something general-purpose AI doesn’t have: context.
- It’s specific to SMBs. SMB inventory planning behaves differently from large-enterprise planning. Netstock’s AI understands the realities of SMB operations, suppliers, lead times and inventory constraints.
- It recommends, not just predicts. Many AI tools stop at the forecast. Netstock’s AI recommends the next action and quantifies its financial impact.
- It can’t be recreated quickly. Any vendor can add a public LLM. Fifteen years of proprietary SMB planning data takes fifteen years to build.
The result so far: Netstock’s AI has made more than 1.5 million recommendations, and over 75% of customers have unlocked $50,000 or more from AI Opportunities.
The numbers underneath still come from deterministic calculations. The same inputs give the same answer every time, so the AI explains and prioritizes figures it didn’t invent.
What do you control in Netstock?
Netstock customers control the data, the planning rules and every decision that reaches their ERP. Netstock’s AI recommends what to do, and your team decides what happens.
| What you control | How |
| The source data | Your ERP stays the system of record. You fix data issues at the source, and Netstock shows which ones to fix first. |
| Planning rules | You set targets such as fill rates, and Netstock calculates safety stock and stocking policies daily to meet them. Lead times can be kept in your ERP or maintained in Netstock. |
| Recommendations | You can review, amend or override any recommended order before it’s placed. |
| What reaches your ERP | Orders reach your ERP only when your team sends them, as pending orders that go through your normal approval workflow. |
| Who sees what | Administrators set each user’s access by location: full, read-only or none. Users never see AI suggestions involving locations they can’t access. |
| What the AI focuses on | Your feedback on AI Opportunities shapes what it surfaces. You can also mute opportunities for a location. |
| Netstock staff access | An app setting lets you allow or block access by Netstock’s Customer Success team, and every visit is logged. |
Where does your data go when the AI processes it?
Your data stays on Netstock’s own infrastructure. Netstock runs its AI models on its own systems rather than sending customer data to general-purpose AI providers such as OpenAI or Anthropic. This approach was chosen in response to customer concerns about how their data is used. Your inventory data is not sent to these external providers to run their models.
Each customer has a separate Netstock instance and database, keeping their data separate from other customers’ data. Data is encrypted when transmitted, and Netstock is ISO 27001 certified, with an independent audit every year.
Netstock also uses pooled, de-identified customer data to train and test its AI models. This helps the models identify patterns across industries and regions that an individual business might not see in its own data. The data is not used to identify your business, and Netstock is contractually committed not to attempt to re-identify it.
How is this different from using ChatGPT or Claude with your ERP data?
General-purpose LLMs are useful for inventory planners. They’re good at one-off analysis, summarizing a planning meeting, and building reporting dashboards. An ERP connected to Claude, for example, can produce a solid view of historical performance. The difference shows up when you need a daily, repeatable process that ends in an order.
| General LLM on raw ERP data | Netstock’s AI | |
| Data quality | Works on whatever the ERP export contains, quirks and all | Validates, cleans and corrects data before planning |
| What it knows | Your past data only, with no behavioral context | Your data plus de-identified patterns from thousands of SMBs across industries and regions |
| How answers are produced | Generated text that can sound confident and still be wrong | Deterministic calculations the AI explains and prioritizes |
| Looking ahead | Mostly reports on what already happened | Recalculates risk daily and flags stockouts and excess before they happen |
| Maintenance | Prompts, exports and custom code someone has to maintain | Built, tested and maintained by Netstock |
| Where your data goes | Depends on the provider’s terms | Stays in Netstock’s own infrastructure |
| End result | An answer in a chat window or a spreadsheet cell | An order ready to approve and send to your ERP |
Can you trust Netstock’s AI with your data?
Yes. Netstock’s AI uses your data for one purpose: to help your team make better inventory decisions. It works from your own ERP data, cleans the data before planning, keeps it in your own Netstock instance, and only learns from it in anonymous, pooled form. Your team stays in control of every decision that reaches your ERP.
