Connecting Priority ERP (פריוריטי) to AI: A Real Integration Guide
Priority is one of Israel's most capable ERPs, and unlike a lot of local business software it ships a genuine, well-documented REST API. That means connecting it to AI isn't a hack — it's straightforward engineering, if you respect the auth model and the edge cases.
If you run Priority (פריוריטי), you're sitting on a structured, authoritative picture of your business: customers, orders, invoices, inventory, purchase orders, financials. The good news for anyone who wants to layer AI on top is that Priority exposes nearly all of it through a real REST API built on the OData protocol. You don't need screen-scraping or a fragile RPA bot — you can read and write Priority data the same way the system does internally.
How Priority actually exposes its data
Priority's REST API maps the system's forms to OData entities. A form like ORDERS or CUSTOMERS becomes an addressable endpoint you can query, filter, page through, and update with standard HTTP verbs. Because it's OData, you get filtering and pagination conventions for free, which matters once you're pulling thousands of records. Authentication comes in three flavors, and choosing the right one is the first real decision:
- Basic authentication — the default; a Priority username and password in the header. Fine for a quick internal job, but you don't want production credentials living in an integration.
- Personal Access Tokens (PAT) — introduced in v19.1, defined in the 'REST Interface Access Tokens' form. You send the token as the username and the literal word PAT as the password. This is the right choice for most integrations: revocable, scoped, and not tied to a human login.
- OAuth2 (Bearer tokens) — the cleanest option, but it requires purchasing Priority's External ID module, so factor in licensing before you design around it.
On top of the API, recent Priority versions ship webhooks. Instead of polling 'is there a new order yet?' every minute, Priority can push an event to your endpoint the moment something changes. That distinction — webhooks versus polling — quietly decides how responsive and how cheap your automation is.
What you can actually build
- Invoice and document intake: an AI model reads incoming supplier invoices or POs (PDF or email), extracts the line items, matches them to the right Priority records, and creates or flags the document — instead of someone keying it in by hand.
- WhatsApp / chat answers grounded in live data: a customer asks 'where's my order?' and an AI agent answers using the real order status pulled from Priority over the API, not a canned reply.
- Reconciliation and anomaly alerts: a webhook fires on each new invoice; AI checks it against expected pricing, quantities and historical patterns, and pings a human only when something looks off.
- Sales and inventory summaries: a nightly job pulls the day's orders and stock levels and an AI writes a plain-language brief — 'three SKUs are about to stock out, two customers are slipping on payment terms' — straight to email or Slack.
- Lead and order triage: incoming leads or orders get scored and routed, with a one-line AI summary written back into a Priority field so the sales team sees context without leaving the system.
No-code or custom code?
Be honest about volume and risk. For a low-volume, low-stakes flow, a no-code tool like Make or Zapier hitting the Priority API can be entirely enough, and I'll happily tell you when that's the answer. But Priority is usually the system of record for money and inventory, and that changes the calculus. The moment you have meaningful volume, data that must stay on your infrastructure, complex mapping between Priority forms and another system, or logic that can corrupt financial records if it gets an edge case wrong, you want real code with tests, retries and idempotency — not a flowchart that silently fails at 2am.
With Priority the API is excellent — the hard part is the business logic. Mapping forms correctly, handling partial failures, and never double-posting a financial record is the actual job.
If you're running Priority and you want to connect it to AI or automate the manual work around it, this is exactly the kind of integration I build — end to end, against the real API, with the reliability your financial data deserves. If you've been looking to hire a developer to connect your existing systems to AI, the contact form below reaches me directly; tell me what Priority is doing today and where it's costing you time, and I'll tell you honestly what's worth automating.
Looking for a developer to connect your systems to AI?
I'm Ariel Gelberg — a senior software engineer and technical partner. I build the integrations and automations that connect your business to AI, end to end.
Let's talk