ALF v2 Migration Schedule
October 31, 2025: ALF v2 officially launches.
November 27, 2025: ALF v1 service will be discontinued, and only ALF v2 will be available.
A Task is a feature designed to handle specific, frequently requested customer work through an automated flow. It now allows for actual work processes to be processed conversationally, going beyond simple question-and-answer exchanges.
For example, if a customer says, "I want to cancel my order," the system does not simply reply, "Check here to see if cancellation is possible." Instead, it can: Look up the order details → Determine if cancellation is possible → Apply the cancellation result to the system → and finally, inform the customer of the final outcome.
This allows support agent to move away from repetitive tasks and focus on core work, enabling the company to efficiently allocate its human resources.
Order Cancellation and Refund Task Example
When a customer requests an order cancellation and refund, the system can check the order list and process the cancellation and refund if the conditions are met.
Reservation Change Task Flow Example
When a customer wants to change a reservation, the system can look up he reservation information and change the reservation in the system as requested by the customer.
Task Examples by Industry
Industry | Task Examples |
|---|---|
Low Involvement E-commerce (Routine/Low-cost Purchases) | Order cancellation, change of delivery address, exchange/return/refund, size/stock check |
High Involvement E-commerce (High-value/Complex Purchases) | Distinguish between free repair and paid repair based on the warranty period, Setting delivery schedule |
Travel/Airlines/Accommodation | Reservation changes, cancellations, and coupon application |
Subscription Service / SaaS | Subscription date change, subscription status inquiry, recurring payment cancellation and refund |
Hospital/Medical | Reservation change, cancellation, and document issuance |
Education / EdTech | Course registration/change/cancellation, textbook delivery status, class schedule/instructor changes |
Mobility Platform | Change/extend/cancel a car rental reservation, handling billing policy questions from customers |
A task consists of one trigger and multiple steps.
Trigger: Defines the condition under which the task will be executed.
Step (Node): Defines what work will be done in each step.
By combining these, you can freely design anything from simple flows to complex business processes.
The trigger is the starting point that defines when the task will begin.
When a customer makes a specific utterance, create the condition to 'Now execute this task (workflow).
1) Trigger Creation
Please write the condition for executing the task in natural language. You can set the task to start automatically when a customer makes a specific request, such as "Please give me a refund" or "I want to cancel my order."
2) Filtering
User Info Filtering: Limit whether the action runs based on customer information (e.g., VIP customers only).
Chat Info Filtering: Limit whether the action runs based on conversation information (e.g., including a specific conversation tag).
Service to run: Restrict the services that will run the action, such as Channel Talk Message, Phone, KakaoTalk, Naver TalkTalk, etc.
Example) If you simultaneously set the trigger phrase to “Refund Request” and the VIP customer filter, the task will only run when a VIP customer requests a refund.
A Step is a small unit that makes up a task. Each Step has a role, like a puzzle piece, and the actual business process is handled following this flow.
This is the Step where ALF processes the conversation according to the instructions. You can define what to say to the customer and the sequence in which the business tasks will be handled. Since you can write the actual support agent's work process in natural language, even non-developers can easily design it.
Writing Instructions
You can simply write a natural language sentence that transfers the actual support agent's work process as is, detailing what ALF needs to do and in what order.
In addition to simple explanations, you can also write down the actual operational rules in sequence.
Example: "After checking the order number to look up the order information, if 30 days have passed, inform the customer that a refund is not possible; otherwise, process the refund."
Agent Steps require a customer's response after executing all instructions to proceed to the next Step. If the customer does not respond, the current Step remains in a pending state, and the process continues to the next Step when the customer sends another message.
Utilizing Variables
When writing instructions, you can use the @ symbol to utilize various variables or conversation handling actions.
1) Reading Variables: User information, chat information, and memory values are available for use.
Example: Replies based on customer information stored in the Channel Talk contact list.
2) Updating Variables: Store new values in user information, chat information, and Memory.
Example: Receive a phone number through a question and save it to the user profile.
3) Running Chat Actions: Automatically handle simple actions such as opening a chat, assigning a team, assigning an agent(assignee), assigning a follower, setting a conversation tag, etc.
What is Memory?
Memory is a temporary storage location that holds information only while a Task is running.
Just as your clipboard remembers text until you paste it elsewhere, Memory temporarily stores information input by the customer (such as an order number, address, or phone number). This information can then be recalled and used when needed in a subsequent Step.
When the Task ends, the information held in Memory disappears as well.
Memory is necessary when you need to reuse information received in a previous step, as shown in the example below.
예시)
The customer enters an order number.
ALF remembers (stores) that order number.
In a subsequent Step, use the order number from Memory to create a branch (conditional logic) or recall it to include it in the final result message.
This Step is for sending a predefined message to the customer.
Please use this when you need to send a predefined message during a Task.
For example, you can show the customer predefined message during the conversation process, such as, "Please wait a moment," or "Your order has been successfully canceled."
You can do more than just display text; you can also add buttons that let the customer click through to a dedicated landing page or run a command.
This Step allows you to directly run JavaScript code in the middle of a Task to perform actual data processing (such as lookup or update), moving beyond a 'simple reply.'
Example: Receive an order number and use an internal order API to check the status → confirm and provide the customer shipping status for that order.
When is it best to use?
When you need to validate values collected in an Agent Step (such as an order number or phone number).
When API communication with internal systems or external services is required.
When you want to use code to evaluate complex conditions and determine the next branch.
When saving values to Memory for reuse in a later Step.
Testing and Debugging
You can immediately test the code you write using the Test button in the top right corner.
The execution results or errors can be viewed in the logs.
This Step automatically runs internal actions necessary for customer support during the conversation with the customer.
Action type | Description |
|---|---|
Set Priority | Automatically classify the chat's priority as 'High,' 'Normal,' or 'Low.' |
Assign / Unassign team | Automatically assign or remove the existing assignment to a specific team. |
Assign / Unassign | Automatically assign or unassign a specific agent (assignee). |
Set / Remove follower(s) | Automatically set or remove a specific agent as a follower of the conversation. |
Set / Delete user tag | Add or remove a tag on the customer profile. |
Add / Delete chat tag | Automatically add or remove a tag on the chat. |
Add chat description | Automatically write a chat description. |
Open a chat | Change the chat status to 'Opened(in progress)' and display it in the Inbox. |
A Task is completed when multiple Steps are connected in sequence to form a complete work process.
There are two main ways to connect Steps:
Go to Single Step (Linear Flow)
Move directly to the next specified Step without any special conditions.
Task Completed: Ends the Task immediately from the current Step.
Move to Single Step: Choose just one next Step to keep the flow moving sequentially.
Branch by condition (Conditional Flow)
You can create multiple paths during a Task, allowing the flow to move to the right next Step based on conditions or the customer's selection.
Branch by Condition: Specify different subsequent Steps based on things like customer(user) information, chat information, Memory values, or operating hours.
Branch by via Button(Customer Response): Allow the customer to select their desired path by clicking a button. (Coming Soon)
You can test whether the Task you set works correctly using the Preview screen.
Please note:
Only the Task settings you are currently viewing can be tested.
The Task operates using the temporarily saved version stored locally (on your PC/browser).
This is the screen where you can view and manage all Tasks in your Channel at a glance.
You can manage similar tasks efficiently by grouping them into folders, while also tracking the execution status of every task.
A Task is only saved locally (on your PC/browser) until you click [Save]. (It remains in a temporary storage state).
When you click the [Save] button, the Task is officially saved and becomes viewable by all agents(team members).
When you [Publish], the Task is deployed and actively used in real conversation with customers.