How to keep control of HR data when using AI

When you connect an AI tool such as ChatGPT to HR Omni, the company decides which integrations to use and every person keeps their own roles and permissions. Your HR data, including employees’ personal data, follows the same access rules that apply in HR Omni, and no new route into it is created.

The assistant acts on behalf of the user signed in with their HR Omni account and does not receive a separate company account. The company chooses the integration, and each user chooses whether and how to connect their account.

Does the AI tool get its own access

No. The AI tool has no user of its own in your company and does not get blanket access to HR data.

When someone asks a question or requests an action, it sends the request on behalf of that person’s HR Omni account. HR Omni sees who the user is, which company they belong to and which roles they hold. If that person cannot see or change something directly in HR Omni, the AI cannot do it for them either. These are the same access controls your company has already configured in the platform.

Who decides whether the connection is enabled

Your company keeps the choice of which connections to use. Integrations are off by default, and HR Omni only offers AI tools it has reviewed and supports.

An authorised company administrator then takes two separate steps:

1. Reviews and approves the disclosure that describes which categories of information and types of action may be available.
2. Enables the specific integration for the company.

Approving the disclosure does not switch the integration on. Enabling one AI tool does not enable another. The company chooses each connection separately.

What each user decides

Enabling the integration for the company does not give anyone personal access automatically. Each user signs in with their own HR Omni account, including through single sign-on with Google or Microsoft where available, and reviews what the assistant is asking to use.

The user can approve the connection, decline it or change the access granted later. That decision only determines which of the capabilities already available to them can be used through AI.

For example, someone may allow ChatGPT to work with leave requests. That does not let them manage other employees’ leave if their HR Omni role does not allow it.

What is checked on every request

The connection is not approved once for any data or action. HR Omni checks every request again before responding.

The checks cover:

1. Whether the user has allowed this type of information or action for the AI tool.
2. Whether the specific integration is enabled for the company.
3. Whether the user still has access to the company.
4. Whether their role allows the requested information or action.
5. Whether the required module is enabled.
6. Whether company settings and the rules for the task allow it.

Only after these checks does HR Omni return information or perform an action. The AI cannot get round a refusal just because the request comes through a conversation rather than the platform’s interface.

Example of a leave request made through AI

Imagine an employee asks AI to submit a leave request. Having a connected account is not enough for the request to be created.

HR Omni checks who the user is, which company the request is for, whether the necessary access has been granted, whether Leave Management is enabled and whether the applicable rules allow the request. If everything passes, the action is recorded in the same process used when working directly in HR Omni.

For practical examples of what people can do in the conversation, read how to use AI for HR tasks every day.

What happens to employee data sent to AI

HR Omni’s checks determine what information can be shared with the AI tool. After that, how it is handled depends on the AI provider’s plan, settings, terms and privacy practices. For ChatGPT, that means your OpenAI plan, settings and terms.

Before enabling an integration, your company should therefore review the provider’s terms as well as the HR Omni disclosure. HR data usually includes employees’ personal data, so this review belongs in your company’s normal data protection and GDPR assessment. Your company decides whether the tool is appropriate for its data and internal policies.

Disabling the integration stops future access through it. It does not automatically delete information the AI provider has already received. More detail is available in our privacy policy.

How access is revoked

The company administrator and each user can stop different parts of the connection.

• The administrator can disable the integration for the whole company. No new requests for company data can then be made through it.
• A user can revoke their own authorisation under Sessions in their HR Omni account settings. The assistant must then ask them to connect again before acting on their behalf.

Disabling the company integration does not necessarily remove every personal authorisation. If the integration is enabled again, an existing authorisation may still be valid. Treat company-level disabling and personal revocation as separate actions.

What is an HR MCP server

Model Context Protocol, or MCP, is an open standard that lets AI tools work with data and actions in other software. An HR MCP server is the part of an HR system that describes, in a form the AI tool understands, which HR data and actions are supported.

The HR Omni MCP server helps the AI model understand the structure of the information available to the user. It does not let the model bypass any of the checks described above. Administrators and employees do not need to understand MCP to decide on access.

What OAuth does

OAuth is a widely used standard for securely connecting an account to another application. Through it, the user signs in to HR Omni and reviews what they are allowing the AI tool to do. Their HR Omni password is never shared with it.

To learn more about the principle behind these checks, read our article on data protection using role-based access control.

How to make an informed choice

When choosing an AI integration for your company, a few practical questions help you use it in a way that fits your organisation:

• Does HR Omni support the AI tool you want to use?
• Which information and actions does the data-sharing disclosure describe?
• Which employees should be able to connect their accounts?
• Are the AI provider’s plan, settings and terms suitable for company data?
• Does your data protection assessment cover using AI with HR information?
• Do users know how to review and revoke their access?
• Do you have internal rules on personal versus company-managed AI accounts?

After that review, you can explore the available integrations or contact our support team with questions about setting things up on the HR Omni side.