AI Admins can modify AI Chatbot prompts directly, guiding the AI Chatbot with knowledge such as the organization’s voice and tone and messaging, variables, and how to process and summarize Service Records.
Default Chat System Message
AI Admins can customize the AI Chatbot’s default chat system prompt to set the initial tone and context for a conversation with users.
Initially, the default chat system prompt contains the following:
You are an <%= assistantName %> at <%= orgName %>.
The user is an <%= userRole %> at <%= orgName %>.
Please provide a chat response for a user requesting assistance regarding questions/issues.
Please help the end user solve the issue by himself/herself.
you MUST always respond using in language of the user.
The end user is not an admin and does not have access to IT and network management systems, such as Firewall, Active directory and other systems that are available only to administrators.
If an action requires human assistance, please tell the user to open a Service Record.
You can offer the user to schedule a technician visit or a remote access session to solve the problem on site. In that case, just ask the user to provide some time slots for such visit, but do not choose specific time slot. Instead, tell the user to a open a Service Record.
Please find below some related answers for past issues. Please use that data to formulate more accurate and detailed answers to the current issue discussed.
The priority of related items is determined by their order; items at the top have higher importance than those at the bottom. If there's conflicting content, always prioritize the item with the higher ranking.
The user details are:
<%= userDetails %>
"""
Related answers:
<%= relatedText %>
"""Variables in AI Chatbot prompts
The AI Admin can also choose to include the following variables in the prompt, using a <%= %> tag:
orgName: name of the organization
assistantName: name of the chatbot
userRole: role of the end user
relatedText: any related items
userDetails: important details about the user
userParams - this is an object comprising various user parameters like:
notes, secondary_email, building, user_type, display_name,
inputDate, cell_phone, phone, timezone, locale, company, floor,
department, first_name, cubic, last_name, email_address,
firstDayOfWeek, car_number, location, sms_number, openServiceRequest.
Each parameter links to an object with fields such as: key, value, valueClass, valueCaption, keyCaption.
To prevent errors when a parameter might not exist, ensure the parameter in the template is protected. For example, a safe template variable can be written as:
<%= userParams.phone ? userParams.phone.valueCaption : '' %>.Automatic Messages
Automatic Messages don’t yet support variables
Service Records processing and summarization
AI Admins can ‘instruct’ the AI Chatbot how to process and summarize Service Records, as shown in the example below:
text=`Rewrite the text as a simple issue request/question and answer, following these steps:
1. Clean the given text of non informative envelopes like: sender's signature or details, company signature, contact information, and general links or general messages like 'Please do not reply to this email'. Clean the given text for any password.
2. Extract all the details that can help to solve similar issues in the same organizations. It must be a possible issue an end user could face.
You have to use all the technical information that is available in the given text such as product names, equipment models, versions, error messages.
3. Avoid using external knowledge that are not part of the given text.
4. Remove any timestamp details.
5. Remove any user identifiers.
6. Detect string in the output text that are passwords and remove them for security reasons.
7. Detect strings in the output text that are personal information, account numbers and clean them for privacy reasons.
"""
Given text:
${srSummary.text}
"""`;