Overview

When an AI assistant is given access to live business data, the natural question for any responsible organisation is: how is that access controlled, and what oversight exists? iVendNext AI Gateway is designed with two principles at its core — the AI can only ever see and do what the authenticated user is already permitted to see and do, and every interaction is logged for administrator review.

This article explains how those principles are implemented in practice.

---

Permission Enforcement — The Foundational Control

The most important security property of iVendNext AI Gateway is also the simplest: the AI assistant operates entirely within the authenticated user's existing iVendNext roles and permissions. It does not have elevated access. It does not bypass module restrictions. It cannot read data the user cannot read, or take actions the user cannot take.

This is not a setting that can be adjusted — it is how the Gateway is architected. Every request the AI assistant sends to iVendNext is sent as the authenticated user. iVendNext applies the same permission checks it would apply if that user were navigating the application directly. If the result would be restricted in the application, it is restricted through the Gateway.

In practical terms, this means:

• A store manager whose iVendNext access is limited to their own store will only receive data from that store, regardless of what they ask.

• A user without access to the purchasing module cannot ask the AI to retrieve purchase order data.

• A user who cannot approve documents in iVendNext cannot instruct the AI to approve them on their behalf.

The AI gateway does not create new access — it provides a more convenient interface to the access the user already has.

How Authentication Works

iVendNext AI Gateway uses OAuth 2.0 with PKCE (Proof Key for Code Exchange) — an industry-standard protocol for secure, delegated authentication. Understanding the basics of this model helps clarify what the AI assistant can and cannot do.

The AI assistant never sees your password. When you connect your AI assistant to iVendNext, authentication happens directly between your browser and iVendNext. You log in with your iVendNext credentials on a screen served by iVendNext itself. The AI assistant receives only a secure access token — a time-limited key — that it uses to make requests on your behalf. Your actual password is never transmitted to or stored by the AI tool.

Access is delegated, not shared. The connection you establish is personal to your account. Another user connecting the same AI assistant to the same iVendNext instance will authenticate with their own account and receive their own scoped token. There is no shared system-level credential that gives the AI elevated or undifferentiated access to the entire iVendNext database.

Tokens can be revoked by an administrator. Only a user with the Administrator role in iVendNext can revoke an AI assistant's access token. Once revoked, the AI assistant loses access immediately and must re-authenticate before it can make requests again. Individual users cannot revoke their own tokens — revocation must be carried out by an administrator. This gives administrators a clear and direct way to manage and terminate AI access when users leave the organisation or when circumstances change.

The Audit Log — Full Visibility of Every Interaction

Every interaction between an AI assistant and iVendNext through the Gateway is recorded in the iVendNext Audit Log. Each log entry captures:

• Who asked — the authenticated iVendNext user whose account made the request

• What was asked — the query or instruction submitted to the Gateway

• What action was taken — the specific iVendNext operation performed (document retrieval, report generation, record update, workflow action, etc.)

• The result — whether the request succeeded, was partially fulfilled, or was denied

This log is maintained within iVendNext itself, not in any external system. Administrators can review it at any time to understand how AI access is being used across the organisation.

The audit log serves several important purposes. It provides accountability — if a record was changed through the AI interface, the log shows who instructed that change and when. It supports compliance — organisations subject to data access regulations can demonstrate that all AI interactions were carried out by authenticated, permissioned users with full traceability. And it provides operational visibility — administrators can see usage patterns, identify unusual activity, and make informed decisions about access policies.

Data Residency — Your Data Stays in iVendNext

A concern that arises frequently in discussions about AI and business data is where the data goes. iVendNext AI Gateway does not extract, copy, or store your iVendNext data externally. When the AI assistant asks a question, the Gateway retrieves the relevant data from your iVendNext instance in real time, returns it to the AI assistant to formulate a response, and that is the end of the transaction.

Your business data does not leave your iVendNext environment in any persistent form. The AI assistant uses the data to answer the question at hand — but it does not build a copy of your database, it does not train on your data, and it does not retain business information between sessions beyond what the AI assistant's own conversation context holds for the duration of that session.

Organisations with strict data residency requirements should review their AI assistant provider's data handling policies — particularly around conversation context — separately from the Gateway's own data model. The Gateway itself introduces no new data residency risk beyond the iVendNext application it connects to.

What Administrators Should Know

For iVendNext system administrators, the following summary covers the key control points for iVendNext AI Gateway:

User-level connections. Each AI assistant connection is tied to an individual iVendNext user account. There is no single service account or API key that grants system-wide access. Access is granted and revoked per user by an administrator — individual users cannot manage their own connection tokens.

Permissions are inherited, not configured. There is no separate permission model to maintain for the Gateway. It inherits and enforces whatever roles and permissions are already set in iVendNext for each user. Keeping those roles current is sufficient to control Gateway access.

The audit log is always on. Logging cannot be disabled. Every Gateway interaction is recorded automatically, giving administrators a complete and uninterrupted record of AI activity on the iVendNext instance.

Revoking access is immediate. If an administrator revokes a user's Gateway access token, the AI assistant loses access in real time. There is no delay, no cache to clear, and no residual access period.

A Note on Responsible Use

iVendNext AI Gateway gives users a fast and accessible interface to business data. Like any powerful tool, it is most effective — and most safely used — when organisations have thought about appropriate use policies alongside the technical access controls.

Organisations that have already defined data access policies for iVendNext will find that those policies translate naturally to the Gateway, because the Gateway enforces the same permission boundaries. For organisations that have not yet formalised data access rules, deploying the Gateway is a good prompt to do so — not because the Gateway creates new risks, but because it makes data access more visible and therefore easier to govern.