Contents

• Functional
• Technical
• Security
• Ciphr roadmap and future capability
• Further information

Functional

What can the AI assistant do?

The first launch has two core capabilities.

• First, employee summaries and data queries - managers and HR professionals can ask questions about employees and their workforce in natural language, getting instant answers on things like team composition, training compliance, absence patterns, skills and qualifications, and pay data.

• Second, holiday booking - employees can check their holiday balance and book time off through conversation, without navigating system screens. Other absence types can also be booked. Manager approvals and rejections will follow in a future release.

Who can use it?

• Any user with access to HR can use the AI assistant, providing their user role has been granted page access to the Chatbot overlay (Ask button).

• What each user can see and do is governed by their existing permissions in the HR system - the AI assistant doesn't create new access, it works within what's already there.

• Employees, line managers, HR professionals, and senior leaders can all use it, but they'll each see and be able to do different things based on their role.

How do we control who can access the AI assistant?

• You can control access from within Role Management.  See AI assistant - managing access guide for more details.

What does Beta mean? 

• The AI assistant is currently available as a Beta product. This means it is fully functional and ready to use, but we are in an active phase of learning - analysing how customers use it in practice to understand what works well and where we can make it even better. That insight directly shapes how we develop and optimise the AI assistant going forward.
• While the product is in Beta, there is no additional charge for access.
   

Does it replace the need to use the HR system directly?

• No - it complements it. Think of it as a faster route to the questions and tasks people ask most often.

• The AI assistant is designed for conversational queries and everyday actions. For more complex configuration, detailed record management, or anything requiring a full system view, users will still work directly in HR.

Will the AI assistant answer questions about how to use HR?

• Not in the current version. The AI assistant is focused on answering questions about your HR data - your people, their records, and actions like booking holiday.

• However, if you do ask a 'how to' guidance question, the assistant will return a direct knowledge base search link (to our customer care portal), generated dynamically from your own wording.

• You can access system guidance directly via the in-app resource centre (? icon)

What languages does it support?

• The AI assistant responds in the language you write in. If you ask a question in Welsh, for example, you'll receive a response in Welsh.

Can it produce charts and visualisations?

• Yes - users can ask the AI assistant to create charts from query results. For example, after retrieving skills and qualifications data across a team, a user can ask for a chart showing the distribution by category. Chart types can be iterated conversationally.

Can charts created by the AI assistant be exported?

• There is no dedicated export or copy button at the moment. However, you can click on a chart image to open it in a new browser tab, then right-click the image and use the browser's 'Save as...' option to save it locally. A more streamlined export option is something we're looking at for a future update.

Can employees book any type of absence through it?

• Yes. Employees can book holiday and other absence types available in your Ciphr HR configuration. 

• The ability to amend or cancel bookings is planned for a future release.

Does absence booking work for employees on a CIPHRFLEXI work pattern?

• Yes. People on a CIPHRFLEXI pattern will be asked for their hours and can add different hours on different days in a single booking: eg 06:00-13:00 Monday, 07:00-14:00 Tuesday.  This will be booked as one record with the correct days and hours.

How does the assistant prevent accidental bookings?

• The AI assistant uses a two-step confirmation flow for anything that changes your data, including booking absence.

• Before submitting a booking, the assistant summarises what's about to happen, validates your entitlement, checks for clashing bookings, and flags if team members are already off the same day. Nothing is submitted to Ciphr HR until you explicitly confirm.

• If you change your mind or want to adjust the dates, the assistant responds to that and re-presents the booking for fresh confirmation - it won't act on ambiguous messages.

Can employees amend or cancel an absence booking through it?

• Not yet. Amendment and cancellation will be added in a future release..

Why doesn't my holiday balance include my pending requests?

• The AI assistant retrieves your holiday balance from the same reporting layer used elsewhere in Ciphr HR. That reporting layer only shows approved bookings, not pending ones, so a request you've submitted but that hasn't been approved yet won't be reflected in the figure.

• The assistant says this clearly when it gives you a balance, so you know what's included. To check your pending requests, see those in Ciphr HR directly.

Can a manager record absence on behalf of a team member?

• Not yet. Currently, only employees can book their own absence through the AI assistant. Managers and HR professionals are not able to record absence on behalf of a team member at this stage.

• The ability for managers and HR professionals to log absence for a direct report - for example, recording sick leave on behalf of a team member - is planned for a future release.

Can it handle follow-up questions in the same conversation?

• Yes. The AI assistant maintains context across a conversation, so users can ask follow-up questions, refine their query, or drill into detail without starting from scratch. Conversation history is also saved for reference, and each conversation is given an auto-generated title based on what was asked. Users can also rename conversation titles to suit them.

What happens when the AI assistant can't answer a question?

• It says so clearly, and directs the user to the appropriate resource - usually HR. It won't guess or generate an answer it isn't confident in. For complex or sensitive situations, it flags that human HR involvement is recommended.

Can line managers approve holiday requests through it?

• Not yet. Manager approvals and rejections are planned for a future release alongside broader absence actions.

Does the AI assistant answer HR policy questions?

• Not yet. HR policy queries are planned for a future release. Employees should continue to refer to policy documents directly in the meantime.

What happens if I misspell an employee's name?

• The AI assistant suggests close matches when the exact name isn't found. For example, if you ask about "Bull Smith", the assistant will suggest "Bill Smith" and ask you to confirm before continuing. This handles common cases like misspellings, partial names, or slight variations in how names are entered.

• If there's no good match, the assistant says so rather than returning data for the wrong person.

Technical

What HR data does the AI assistant have access to?

The AI assistant can access the following data areas within HR:

• Available now: allowances; absence and holiday entitlements; current and historic job and pay data; development needs; employee directory; payments; personal data; skills and qualifications; and who is off (team absence visibility).

• Coming soon: objectives.

• Examples of areas not included yet: absence verification; additional information fields; case management; safeguarding records; policy acceptance; and training records. 
  This is not an exhaustive list - the Beta is focused on the available data areas listed above, and further areas will be added over time.

How does it connect to our HR data?

• The AI assistant connects to HR through Model Context Protocol (MCP) servers, which link to the HR application via its report service. Because it uses the existing report service, no new permissions were created - the AI assistant works within the same user permission model already in place in HR. Each user sees exactly what their role permits, nothing more.

• See AI assistant - how it works for more details.

What is Model Context Protocol (MCP)?

• Model Context Protocol (MCP) is an open standard that allows AI models to connect securely to external tools, databases, and data sources. Think of it as a purpose-built bridge between an AI assistant and enterprise systems like Ciphr HR - it controls what data the AI can access and how, without requiring new permission structures to be built from scratch.

Does it work in real time, or is the data cached?

• The AI assistant retrieves data in real time, connecting directly to the HR report service each time a query is made. This means responses always reflect the current state of your HR data.

What are the system requirements to use it?

• The AI assistant is accessed through HR via the Ask button in the top navigation bar. No additional software, plugin, or device is required. It works in any browser supported by HR.  

• Access to the Ask button is controlled by user role page access so you can decide which users can interact with it.

How long are conversations kept?

• Conversations are retained for 3 months. After that, they're automatically removed and won't appear in your chat list. They are also permanently removed from our data storage and infrastructure. 

• If you want to keep a record of a specific conversation, copy the content out before it ages off.

Is the AI assistant available on the HR mobile version?

• The AI assistant is available on the Ciphr HR mobile front-end, with the same 'Ask' button and the same capabilities as the desktop experience.

Is there an API or integration available for the AI assistant?

• Not yet. Our longer-term roadmap includes exploring an API for Ciphr AI solutions, which would allow customers to connect the AI assistant to their wider technology ecosystems.

What AI model powers it?

• We continually optimise and refine our AI infrastructure, which means the underlying models may change over time. Current details are available from the Ciphr Trust Centre.

Does the AI assistant use a large language model (LLM)? What does that mean for the responses?

• Yes - the AI assistant is powered by a large language model (LLM), which is what enables it to understand natural language questions and respond in plain, conversational English rather than returning raw data or requiring specific search terms.

• It is important to understand what the LLM does and doesn't do:

  • When a user asks a data question - "Who in my team hasn't completed GDPR training?" - the LLM interprets the question and retrieves the answer from your HR data. The response reflects what is actually in the system.

  • For more general guidance - explaining how a process works, or suggesting a sensible next step - the LLM draws on its broader training. This is where its knowledge goes beyond what is stored in your HR system.

  • Ciphr's AI assistant is designed to be clear about the difference: factual responses from your data are attributed to their source, and the AI assistant will indicate when it cannot answer from system data directly.

• As with any AI-powered product, outputs should be treated as a helpful starting point rather than a definitive authority - particularly for complex or sensitive HR matters, where human judgement remains essential.

What happens if the AI assistant is unavailable or returns an error?

• If you receive an unexpected response or no response at all, try submitting your prompt again.

• If the issue persists, starting a new chat will clear the session and often resolves the problem.

• If the AI assistant remains unavailable, please raise a support ticket with the Ciphr Customer Care team.

Security

For any security related questions, please refer to the Ciphr Trust Centre.

Ciphr roadmap and future capability

What's coming next for the AI assistant?

The immediate focus for the AI assistant is in these areas:

• Expanding holiday and absence capability is next, including the ability to amend and cancel bookings, bulk approvals for managers, return-to-work interview support, and approval delegation.

• HR policy queries, allowing employees and managers to ask questions about HR policies with cited answers, is also planned.

• Improved UX including dynamic suggested prompts tailored to each user type (so employees, managers and HR professionals each see prompts relevant to them). We're also exploring saved prompts, movable chatbot panel positioning, and chart export.

What else is on the Ciphr roadmap for the AI assistant?

• Beyond the immediate focus, several further areas are in planning. Employees will be able to manage their own personal details through conversation - updating address, bank information, job-related documents, and more. Payroll integration is also in planning, with two initial use cases: anomaly detection to help payroll teams spot errors before they become problems, and a payslip explainer so employees can understand their payslip through conversation.

• Further ahead, we are exploring integration with Ciphr LMS - enabling the AI assistant to guide employees on career development and learning opportunities based on their performance and talent data - and making the AI assistant available within Microsoft Teams, so employees and managers can access it without leaving the tools they already use every day, and the ability for the AI assistant to answer questions about how to use Ciphr HR itself, helping users get the most from the system without needing to contact support.

Can customers influence what gets built next?

• Yes. As with all Ciphr products, customers can provide product feedback by raising a support ticket with the Ciphr Customer Care team. Simply using the AI assistant also helps - the more it's used, the more we learn about what's working and what to improve.

Will the AI assistant eventually be able to take more actions autonomously - without being asked?

• That's the direction of travel. The current product is conversational - it responds to questions and requests. Future agentic capability, where the AI proactively surfaces insights or takes actions in the background, is part of the longer-term vision.
   

Further information

• AI assistant - managing access
• AI assistant - how it works
• AI assistant - tips and tricks
• AI assistant - capabilities
• Get more from your AI assistant - suggested prompts
• How to raise a ticket through the Customer Care Portal
• Ciphr Trust Centre