Connecting Comax (קומקס) Retail ERP to AI and Automations

Comax (קומקס) is the operational spine for a lot of Israeli retail and wholesale: it holds the product catalog, the real-time inventory per branch, the purchase orders to suppliers, the customer and loyalty data, and — through its POS — the sales themselves. If you run a chain, a warehouse, or a store with an online shop bolted on, Comax is usually where the truth lives. The reason engineers want to connect it to AI is that every one of those records is a structured business event: this SKU sold at this branch at this price, stock dropped below threshold, this supplier is late, this customer hasn't bought in ninety days. That is high-signal data — and in most companies it sits locked inside Comax, queried only when someone manually opens a report.

How Comax actually exposes data

Comax is an established Israeli ERP, not a modern API-first SaaS, so the integration surface is a web service layer rather than a fashionable REST playground — but it is real and it works. The practical building blocks you assemble are:

• Authentication — you get integration credentials (a login/key) from Comax or your reseller and authenticate against the web service endpoint. Credentials and endpoints are environment-specific, so a real integration keeps them in config/secrets, never hardcoded.
• Catalog / items — read and update products (פריטים): SKUs, barcodes, prices, categories, suppliers. This is the dimension table everything else joins to.
• Inventory — query stock levels per branch/warehouse (מלאי לפי סניף). This is the data behind every reorder, transfer and out-of-stock alert.
• Orders & documents — pull sales, purchase orders and delivery documents, and in many setups push orders back in (e.g. an online order flowing into Comax).
• Customers — read and update customer records (לקוחות) and loyalty data so AI matches activity to the right account instead of creating duplicates.
• Branches — because Comax is multi-branch by design, almost every query is scoped by branch, and your data model has to respect that from day one.

Because Comax leans on request/response web services rather than push webhooks, most integrations run on a scheduled sync — pull the deltas every few minutes — plus on-demand calls when a flow needs fresh data. That is a perfectly solid pattern; you just design for it deliberately instead of pretending real-time events exist when they don't.

What you can actually build

1. AI demand forecasting and reorder: a job reads sales velocity and stock per branch, predicts run-out dates, and drafts purchase orders — so you reorder before the shelf is empty, not after.
2. Out-of-stock and dead-stock alerts: AI watches inventory across branches and pings the right manager in Hebrew on WhatsApp when a top seller is running low, or when stock has sat untouched for months and needs a markdown.
3. Online order to Comax, automatically: an order from your Shopify/WooCommerce/Wix store is created in Comax with the right items, customer and branch — no manual re-keying, no end-of-day batch.
4. Plain-language retail answers: a chatbot that answers "which SKUs sold best in the Tel Aviv branch last week, and which suppliers are behind on deliveries?" by querying Comax and summarizing — no report export, no analyst round-trip.
5. Smart customer reactivation: AI spots customers who lapsed, drafts an on-brand message with a relevant offer based on their purchase history, and routes it to your messaging channel.

Where the real work is

The AI model is the commodity now; the engineering lives in the seams. The hard parts with Comax are the unglamorous ones: reconciling the scheduled sync so you never double-count a sale or miss a delta when a job fails mid-run, making any write-back (orders, stock adjustments) idempotent so a retried call doesn't duplicate a document, mapping Comax's item and branch codes onto your own model, handling multi-branch scoping correctly so an alert fires for the right store, respecting that inventory is a moving number you read at a point in time, and keeping customer and pricing data out of any AI prompt or log it shouldn't reach. This is operational infrastructure for a business that loses real money when stock or orders are wrong — so quiet correctness matters far more than a clever model.

With retail ERP the AI is the dashboard everyone sees. The boring, idempotent sync that keeps inventory and orders correct across every branch is the engine — and the engine is the part that has to be right.

No-code or custom code?

For a single low-stakes flow — posting a daily sales total to a Slack channel, say — a no-code tool like Make or Zapier can sometimes bridge the gap, and I'll tell you honestly when it can. But Comax's web service layer, multi-branch scoping and write-backs push you toward custom code fast: scheduled deltas that never lose data, idempotent order and inventory writes, correct branch handling, customer data that stays off third-party platforms, and AI logic too specific for any pre-built connector. If you're looking to hire a developer to connect Comax (קומקס) to your existing systems and AI-driven automations, this is exactly the work I do — building the secure, reliable layer between your retail ERP and AI, end to end. The contact form on this page reaches me directly; tell me what your inventory, orders and customers should be doing on their own, and I'll build it.