Conversational AI on Oracle EBS — WhatsApp & Mobile for Enterprise Teams
How we built a zero-disruption AI layer on Oracle E-Business Suite that lets enterprise users query live ERP data — PO status, invoices, inventory — via WhatsApp and mobile, without touching core ERP logic.
The Challenge
A large manufacturing enterprise running Oracle EBS faced a critical productivity gap: field teams, procurement officers, and finance staff had no easy way to access live ERP data on the go. Checking PO status, invoice approvals, or stock availability required logging into Oracle — a complex, desktop-bound process unavailable in the field. This created daily delays, bottlenecks, and dependency on back-office staff to relay basic ERP information.
- ✕Field and sales teams could not access Oracle EBS data from mobile or WhatsApp
- ✕Every PO status or invoice query required a back-office person to log into Oracle and report back manually
- ✕No proactive alerts when ERP events occurred — teams reacted late to stock-outs, overdue invoices, or PO delays
- ✕Exposing Oracle EBS APIs directly to mobile clients was a security and compliance risk
- ✕Business needed real-time answers without any changes to core Oracle EBS configuration or logic
Solution Architecture
How It Works — Step by Step
User Sends a Request
A field user or enterprise staff sends a natural-language message via WhatsApp or the Mobile App — e.g. "What is the status of PO 4500123?" or "Check invoice INV-2024-009".
LLM Understands Intent & Extracts Entities
The AI backend uses an LLM to detect the user's intent (PO status, invoice check, stock query) and extract key ERP entities like PO number, invoice ID, or product code — no structured input required.
Orchestration Applies Business Rules
The Wrapper API & Orchestration Service validates the user's access permissions, applies business rules, and determines which Oracle EBS module to query — Procurement, Finance, Inventory, or Customer.
Secure Read from Oracle EBS
The Integration Layer queries Oracle EBS via secure, read-only APIs and views — fetching live data from Procurement, Accounts Payable, Inventory, or Order Management. No ERP core logic is changed or exposed.
LLM Formats a Business-Friendly Response
Oracle EBS returns structured ERP data. The LLM converts raw data into a clear, human-readable response — "Your PO 4500123 is approved and dispatched. Expected delivery: 12 Apr 2024."
Proactive Alerts & Notifications
The same system supports proactive push alerts triggered by Oracle EBS events — overdue invoices, low stock, delayed shipments — delivered directly to WhatsApp or mobile without any manual monitoring.
Zero Changes to Core Oracle EBS
The entire AI layer sits outside Oracle EBS. All queries use read-only views and APIs. No ERP configuration, schema, or business logic is modified — making this safe, compliant, and fully reversible.
User (WhatsApp / Mobile App) → Secure Message Gateway → LLM Intent & Entity Extraction → Wrapper API & Orchestration (Business Rules + Access Control) → Integration Layer (Read-only Oracle EBS APIs/Views) → Oracle EBS Modules (Procurement, Finance, Inventory, Customer) → Response Normalisation → LLM Business-Friendly Formatting → Delivered back to User. Proactive alerts flow: Oracle EBS Events → Event Trigger Engine → LLM Formatting → Push Notification to WhatsApp / Mobile.
Client Testimonial
Our field teams can now check PO status and stock availability on WhatsApp in seconds — without calling the back office. Zgrow built this on top of our existing Oracle EBS without changing a single line of core ERP configuration. It's been transformational for our operations.
🛠 Tech Stack Used
Want Similar Results?
Let's discuss how we can deliver this for your business.
Book Free Consultation