> For the complete documentation index, see [llms.txt](https://en.help.firstline.cc/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://en.help.firstline.cc/feature/ai-assistant.md). # AI Assistant {% hint style="info" icon="eyes" %} The AI assistant is expected to go live in **Q3 2026** Official launch.\ The features and workflows described on this page are planned features. The actual launch version may differ due to product development, test results, and feature adjustments. Supported data scopes, analytics metrics, operating methods, and UI presentation are all subject to the official launch version. Please refer to subsequent product update announcements. {% endhint %} ### Overview {% columns %} {% column width="16.666666666666664%" %} {% endcolumn %} {% column width="83.33333333333334%" %} The AI assistant is an intelligent work assistant integrated into the FIRST LINE work platform, helping service staff quickly find information, understand customer context, check ticket status, and analyze operational data. Unlike general AI chat tools that can only answer questions, the AI assistant can query FIRST LINE's knowledge base, customers, tickets, service records, and operational data according to the current user's permissions, turning work that originally required switching between multiple pages and searching item by item into natural-language Q\&A. {% endcolumn %} {% endcolumns %} For example, you can directly ask: * “What is the return/exchange policy?” * “Help me find Wang Xiaoming’s customer record” * “What unresolved tickets do I still have?” * “Is it business hours now?” * “Please count the daily CSAT trend for the last 30 days and present it as a line chart.” **The AI assistant determines the information source to query based on the question and organizes the results into an easy-to-read format**; some statistical questions can also generate charts directly, reducing the time needed for manual searching, consolidation, and cross-checking of data. If you want the AI assistant to perform data statistics or generate charts, it is recommended to use descriptions such as “count,” “analyze,” or “compare” directly in the question, and include the time range and analysis items. This can reduce the chance of the question being interpreted as a general knowledge query. {% hint style="info" %} The data obtained by the AI assistant is still subject to FIRST LINE's original permission and data visibility restrictions. Users cannot use the AI assistant to query data they do not already have permission to access. {% endhint %} *** ### How to use the AI assistant After logging in to FIRST LINE, you can find**the “FIRST LINE AI Assistant” F-shaped icon above the navigation bar**. Click the icon to open the AI assistant and directly enter the question you want to query in natural language, without having to remember specific commands or syntax. For example: **Query product instructions** “How do I set up business hours for live chat?” **Search for customers** “Help me find the customer with phone number 0912-345-678.” **Keep track of pending work** “What unresolved tickets do I currently have?” **Analyze service status** “Please count the service interaction volume for each day of the last 30 days and show the trend with a line chart.” The clearer the question, the more accurately the AI assistant can usually determine the query scope. If the question involves a specific period, customer, ticket status, or analysis criteria, it is recommended to state them directly in the question.
*** ## Things you can currently do ### Query the knowledge base, SOPs, and product documents When you need to confirm product instructions, company policies, service procedures, or how to respond to customers, you can directly ask the AI assistant. The AI assistant can search the knowledge content available in FIRST LINE, including the knowledge base, product documents, and other established knowledge files. For example: * “How should the refund process be handled?” * “What is the service SOP for VIP customers?” * “How do I set up this feature?” * “How should I reply when a customer asks to cancel an order?” If the summary found is not enough to answer the question, the AI assistant can further read the relevant document content and answer based on the complete information. *** ### Search and view customer information When you need to confirm a specific customer, you can directly provide the known customer information and let the AI assistant help search. The currently available search criteria include: * Name * Phone * Email * Member ID * ID number For example: “Help me find Wang Xiaoming.” Or: “Check the customer with member ID A123456.” After finding the customer, you can also continue asking for detailed information about that customer, including basic information, contact details, address, custom fields, do-not-disturb settings, upcoming appointments, and recent tickets and service records. For example: “What tickets have they had recently?” “When is this customer's next appointment?” The AI assistant will only search within the range of data visible to the current user. *** ### Search and organize tickets The AI assistant can help search tickets by ticket status, priority, or keywords. By default, the AI assistant will prioritize tickets currently assigned to you; if you need to search other tickets within your permission scope, you can state that directly in the question. For example: * “What unresolved tickets do I still have?” * “Find high-priority tickets that are not yet closed.” * “Have there been any tickets related to damaged packaging recently?” * “Help me find all refund-related tickets within the currently visible scope.” This kind of query is suitable for quickly organizing to-dos or locating specific issues among a large number of cases. *** ### Check business hours When you need to confirm whether it is currently service hours, you can directly ask the AI assistant. For example: * “Are we open now?” * “What time do we close today?” * “Are we open tomorrow afternoon?” * “When is the next business opening?” The AI assistant can confirm whether the business is currently open, the opening and closing hours for a specified date, and the next time business will start. *** ### Find suitable colleagues and reassignment targets When handling cases that require help from other specialists, you can use the AI assistant to search for colleagues in the team with specific skills. For example: * “Who is familiar with refund issues?” * “Who can this technical issue be transferred to?” * “Which colleagues have VIP customer service skills?” The AI assistant will help find suitable people based on the skills and proficiency levels configured in the system. If you're not sure what skills are currently available, you can also ask directly: “What specialist skills are currently available?” *** ### View past actual conversation content In addition to viewing service records, the AI assistant can also further read the actual conversation content within the service records. For example: * “What did this customer actually say last time?” * “What did we promise the customer on the last call?” * “Help me summarize the key points from the last service record.” This feature is suitable for quickly restoring the service context before continuing to handle a case, reducing repeated questions to customers due to incomplete handoff information. *** ### Statistics and chart generation The AI assistant can directly count service and ticket data in FIRST LINE and generate statistical results or charts based on the question. You can specify a time range, specialist, channel, or other conditions, for example asking “Please count the service interaction volume for each day of the last 30 days,” or further cross-analyzing “Please analyze the service interaction distribution by weekday and time period over the last 30 days.” {% hint style="info" %} **If the goal is to analyze actual operational data in FIRST LINE, it is recommended not to ask overly short or conceptual questions.** For example, “How long do customers usually wait?” may also be interpreted as a general customer service knowledge question. If you want the system's waiting-time statistics and charts, you can ask instead: **“Please count the distribution of customer waiting times for agents over the last 30 days, and use a chart to show the proportion of each waiting interval.”** It is recommended that the question include at least:**The metric to be counted, the time range, and the way you want it compared or presented**. {% endhint %}
The following analyses are currently supported: | Analysis item | What you can learn | Example question | | ------------------------------------------------ | --------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Ticket status distribution** | The number and distribution of each ticket status | “Please count the status distribution of currently visible tickets and list the number of each status.” | | **Daily service interaction trend** | The daily number of customer interactions and trend changes | “Please count the service interaction volume for each day of the last 30 days and present it as a line chart.” | | **Service channel distribution** | Interaction volume and share across different service channels | “Please count the interaction volume and share of each service channel for the last 30 days.” | | **Service tag distribution** | The number and distribution of interactions corresponding to each service tag | “Please count the number and share of interactions for each service tag this month.” | | **Service result distribution** | The number and distribution of interactions across different service results | “Please count the number and distribution of interactions for each service result over the last 30 days.” | | **Time period service interaction distribution** | Interaction volume across different time periods in a day | “Please count the service interaction volume for each time period over the last 30 days, and identify the time period with the most interactions.” | | **Weekday service interaction distribution** | Interaction volume across different days of the week | “Please count the service interaction volume for each weekday over the last 30 days, and identify the weekday with the highest interaction volume.” | | **Weekly service peak distribution** | Cross-compare weekdays and time periods to identify weekly service peaks | “Please analyze the service interaction distribution by weekday and time period over the last 30 days, identify the main peak periods, and present them in a chart.” | | **Daily average handling time** | The average handling time for completed interactions each day | “Please count the daily average handling time for the last 30 days and show the trend.” | | **Daily average first response time** | After a customer starts interacting, the average time each day until an agent responds | “Please count the daily average first response time for this month and show the trend with a line chart.” | | **Message inbound/outbound distribution** | Distribution of the number of incoming customer messages and outgoing agent messages | “Please count the number and share of incoming customer messages and outgoing agent messages over the last 30 days.” | | **Daily CSAT trend** | The change in average customer satisfaction score over time | “Please count the daily average CSAT score for the last three months and show the trend.” | | **CSAT score distribution** | The number and share of different satisfaction scores | “Please count the number and share of each CSAT score this month.” | | **Daily abandoned service trend** | The change in the number of interactions abandoned before being served by an agent | “Please count the number of abandoned services each day over the last 30 days and show the trend.” | | **Waiting time distribution** | Distribution of the time customers wait for an agent to respond | “Please count the distribution of customer waiting times for agents over the last 30 days, and use a chart to show the proportion of each waiting interval.” | | **Daily transfer trend** | The number and change in interactions handled by multiple agents | “Please count the number of transferred services each day over the last three months and show the trend.” | | **Overdue ticket distribution** | The distribution of tickets that are not yet closed and have exceeded the due date | “Please count the currently overdue and still open tickets, and show the distribution by status.” | | **Daily average after-call work time** | The average time required for service handling | “Please count the daily average after-call work time for the last 30 days and show the trend.” | | **Daily idle timeout trend** | Compare the number of occurrences of long customer inactivity and long agent inactivity | “Please compare the number of customer idle timeouts and agent idle timeouts each day over the last 30 days, and present it in a chart.” | | **Ticket category distribution** | The number of cases in each ticket category | “Please count the number and share of cases in each ticket category this month.” | | **Ticket sub-status distribution** | The distribution of cases across different ticket sub-statuses | “Please count the number and distribution of cases in each ticket sub-status over the last 30 days.” | | **Daily AI credit usage** | Daily AI feature credit usage and changes | “Please count the AI credit consumption for each day over the last 30 days and show the trend.” | {% hint style="info" %} “Service tag” is classification information at the service interaction level; “service result” records the final handling result of the service; “ticket sub-status” is a finer status breakdown at the ticket level. The three are used differently, so when analyzing, it is recommended to choose the corresponding item based on the actual process you want to understand. {% endhint %} #### Cross-analysis In addition to single metrics, you can also cross-analyze across different dimensions to help identify where service volume is concentrated among staff, channels, service tags, and time periods. | Analysis method | What you can learn | Example question | | ------------------------------------------------- | ---------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Agent time-period service volume distribution** | Compare service interaction volume across different agents in each time period | “Please analyze the service interaction volume of different agents in each time period over the last 30 days, and identify each agent's main service periods.” | | **Channel weekday service volume distribution** | Compare the interaction volume of different service channels across each day of the week | “Please compare the service interaction volume of LINE, phone, and live chat on each weekday over the last 30 days.” | | **Service tag time-period distribution** | Compare the interaction volume of different service tags across time periods | “Please analyze the interaction volume of the 'refund' service tag across time periods over the last 30 days and identify the main concentration periods.” | | **Weekly service peak distribution** | Cross-compare overall interaction volume by weekday and time period | “Please analyze the service interaction distribution by weekday and time period over the last 30 days, and identify the weekly peak service periods.” | #### How to ask If you want the AI assistant to perform statistics or generate charts, it is recommended to provide the following in the question at the same time: 1. **Analysis item**The metric you want to count, such as waiting time, interaction volume, CSAT, or ticket status. 2. **Time range**: for example, the last 7 days, the last 30 days, this month, or the last three months. 3. **Filter or comparison criteria**: for example, a specific agent, channel, tag, or ticket category. 4. **Presentation format**: if you want to view a chart, you can directly request “present as a chart,” “compare trends,” or “list the share of each interval.” For example: “Please count the service interaction volume for each day of the last 30 days and present it as a line chart.” “Please compare the interaction volume and share of each service channel over the last three months.” “Please analyze the service interaction distribution by weekday and time period over the last 30 days, and identify the weekly service peak.” “Please compare the interaction volume of different service tags across time periods over the last 30 days.” “Please count the number and share of cases in each ticket category over the last 30 days.” {% hint style="warning" %} If a question can be interpreted as both “check documents” and “check actual data,” it is recommended to state directly that you want to**count the actual data in FIRST LINE**. For example: * More vague: “How long do customers usually wait?” * Suggested: “Please count the distribution of customer waiting times in FIRST LINE for the last 30 days and present it in a chart.” * More vague: “Which day of the week is busiest?” * Suggested: “Please count the service interaction volume for each weekday in FIRST LINE over the last 30 days, and identify the weekday with the highest interaction volume.” {% endhint %} The AI assistant will perform statistics according to the specified conditions; analytical results that support chart presentation will also generate charts directly in the response, making it easy to quickly interpret trends and distributions. *** ### Usage tips The AI assistant uses natural-language operation and does not require learning additional query syntax. However, providing more complete conditions usually yields more accurate results. For example, instead of: “Help me check tickets.” It is recommended to ask: **“Help me find the high-priority tickets currently assigned to me that are not yet closed.”** When analyzing data, it is also recommended to provide a clear time range, analysis item, and expected presentation format, for example: **“Please count the interaction volume and share of each service channel over the last 30 days and present it in a chart.”** If the question itself is very short and could also be a general knowledge question, such as “How long do I need to wait?”, “Which time period is busiest?”, or “Which weekday has the most?”, it is recommended to add descriptions like “count data from the last 30 days” or “analyze service interaction volume” so the AI assistant can more easily determine that you want actual operational data rather than knowledge-base content. If the question involves a specific customer, it is recommended to first provide identifiable information such as name, phone, email, or member ID, and then continue asking about that customer's tickets or service records. *** ### Notes The AI assistant's answers are generated based on the data currently available in FIRST LINE, so the completeness of the results still depends on the existing data and settings in the system. Please note when using: * The AI assistant follows the current logged-in user's data permissions and visibility scope. * If customers, tickets, or service records have not yet been created in the system, the AI assistant cannot retrieve the relevant information. * The statistical results are calculated according to the specified conditions and the current system data. * For policies, commitments, contracts, or other important decisions, it is still recommended to confirm the original documents and actual data before proceeding with further actions.