The AI assistant can personalize responses using information about the employee, the organization, the conversation, and connected systems.
This helps the assistant provide responses that are:
- Relevant to the employee, based on information such as country, role, and what they have already shared.
- Consistent with organizational content, including approved policies and other knowledge.
- Aligned with the organization’s tone of voice, while still following policy, security, and privacy requirements.
Personalization does not mean the assistant creates policies or makes assumptions about information it lacks. If approved knowledge or a connected system does not provide the information needed, the assistant should say so.
Employees must be signed in to receive personalized responses. Unsigned visitors see a sign-in prompt instead.
What the AI assistant uses
Employee profile
When an employee signs in, the assistant can use selected profile information to provide relevant responses. This can include:
- Name
- Country
- Job title
- Department or business unit, where available
The assistant uses this information to personalize responses and help identify relevant knowledge or tools.
For example, an employee's country can help the assistant identify the appropriate country-specific policy.
Profile information does not provide access to the employee's entire HR record. Live information, such as a current leave balance, comes from a connected system rather than from profile information.
The assistant is designed to use only the information it needs for the request. It should not ask employees for information already available to it or request sensitive information, such as passwords, full bank details, or government identification numbers.
Saved preferences
The assistant can remember information that an employee clearly provides as a lasting preference, such as a preferred name or working preference.
Saved preferences help personalize future conversations. They are limited to the information the employee provided and can be updated as preferences change.
Organization voice and branding
Organizations can define how the assistant communicates through their configured theme.
This can include:
- Company or brand name
- Tone of voice, such as formal or friendly
The assistant follows the configured tone unless it conflicts with legal, security, or HR policy requirements. Policy and security requirements take precedence.
Conversation context
The assistant uses the current conversation to keep follow-up questions and requests relevant.
It can use information from earlier messages in the conversation, so employees do not need to repeat information unnecessarily.
For example, if an employee asks about their leave entitlement and then asks, “What about next Friday?”, the assistant can use the context from the earlier question when responding.
Knowledge
For organization-specific questions, the assistant searches the indexed knowledge available to it. This can include:
- Knowledge Management articles
- Uploaded knowledge documents
- Connected SharePoint content
The assistant bases policy answers on the content it retrieves and provides sources where the channel supports them.
Knowledge can also be targeted using information such as country and audience. This helps the assistant identify content intended for the employee, such as a manager-specific or country-specific policy.
These settings help the assistant find relevant content, but they are not a substitute for appropriate access controls.
Files that an employee attaches to a conversation can be used during that conversation. They are not added to the shared knowledge library.
Live systems
Connected tools can provide current information about an employee.
For example, a Workday tool can retrieve an employee's current leave balance. This information comes from the connected system rather than from the employee profile.
The assistant can also use configured tools to complete supported actions, such as submitting a time-off request.
Responses across channels
The assistant adapts its responses to the channel employees are using.
| Channel | How responses are presented |
| Employee portal and mobile | Richer formatting, sources, and interactive visuals when available |
| Microsoft Teams and Slack | Concise chat responses, with sources as links or named policies and visuals presented as images |
| Outlook | Email-appropriate responses that focus on the latest question in the conversation |
The assistant adapts formatting and response length to the channel while using the same underlying capabilities and configuration.
Language
The assistant responds in the same language as the employee's latest message.
It continues using that language unless the employee clearly switches to another language. The assistant does not automatically translate content unless translation is requested or provided through a configured capability.
Date and time
The assistant has access to the current date and time when responding.
This allows it to interpret requests that depend on dates, such as “today,” “this month,” or “this policy year,” without the employee needing to provide the exact date.
What personalization does not mean
Personalization does not mean:
- The assistant receives a complete copy of the employee's HR record.
- The assistant can access another employee's private information.
- The assistant can use profile information to guess missing HR data.
- The assistant can bypass permissions or other access controls.
- The assistant can skip confirmation when a configured action requires it.
- The assistant can provide personalized responses to employees who are not signed in.
Personalization helps the assistant use the right context. It does not expand what the employee is allowed to access or what the assistant is configured to do.