Create your AI Bot scenario
Table of Contents
Available for AlcmeonChat, WhatsApp, LINE channels
Using generative AI, Alcmeon bots can generate near-human responses to customer queries, providing accurate information and reducing the need for advisor intervention.
To produce these responses, the AI bot evaluates the context of the conversation and uses your knowledge base to provide accurate and relevant answers.
Create your AI Agents
In the “Alcmeon Settings / AI” section you have the option to configure the AI Agent(s) that will intervene in your bot scenario.
For more information, see our detailed article: Creating your AI Agent .
Setting up your AI Bot scenario
Here's the classic chain of steps that will allow you to integrate AI responses into your Alcmeon bot scenario, which we'll detail:
1. Retrieve the customer's question
The first part of this scenario branch is to ask the client to ask a question, using 2 special steps that will allow this question typed by the client to be transmitted to the AI.
Use the “Subbot” step, and click the “pencil” icon to edit it. Choose the “internal” subbot called “Capture message content” and validate. The step that then appears in your scenario has 3 outputs:
- Message content captured : This step is successful, the “message_content” variable is now set to the user’s last message. Connect this output to an “Alcmeon Action” step (see below)
- Message format not supported : The user's last message contains an unsupported format, such as a Line sticker or an image or video. We recommend linking this output to a step that asks the customer to type a new question.
- Exit failed : A technical error occurred. We recommend linking this exit to a step that asks the customer to type a new question.
Connect the "Message content captured" output to an action step in which you configure the action "Save the contents of the message_content variable". This will allow this variable to be passed to the AI Connector.
Connect the output of this action step to the optional next step, or if you don't want to implement it to the AI Connector step.
2. (optional) Introduce the AI assistant
This step is optional, but we recommend it. It lets the customer know that an artificial intelligence will now respond, and that this may take a few seconds, which is often the case with generative AI.
Connect the output of this step to the IA Connector step.
3. Hand over to AI
This step will transmit the customer's question to your AI Agent, who will then find an answer or indicate that they were unable to answer, for several reasons that we will detail.
Configure the AI Connector
.png)
Click on the “pencil” icon for this step to open its settings screen. You will find 3 settings there:
Maximum number of characters for a message : Allows you to indicate to your AI Agent the maximum number of characters it should use to formulate its response to the customer, which must not exceed the limit imposed by the channel such as WhatsApp for example. Note: Generative AIs sometimes do not respect this type of indication exactly; if exceeded, Alcmeon will split the AI response into 2 messages.
String that prefixes all AI responses : Allows you to prefix all AI responses with the same string. Our tip: simply use an emoji + a space to let your customer know they're talking to an AI bot.
AI Agent : From this drop-down menu, choose the Agent that should respond. This is configured in the “Settings / Alcmeon AI” section.
Management of the different outputs
Below for each exit you will find a description of it, the bot event (1) that fires when that exit is taken, and our recommendation on how to link that exit to the next step.
Exit Name |
Description |
Bot event |
Recommendation |
|---|---|---|---|
AI - response (no advisor needed) |
The AI responded with a valid response which was sent to the client. | AI Event: AI computed an answer |
Link to an “AI Satisfaction” step |
AI - no response (off topic) |
Irrelevant message (subject outside the customer's scope, problematic messages, etc.). | AI Event: AI has no answer for you |
Transfer to an advisor |
AI - response (advisor needed) |
The AI bot sent a response but indicated that an advisor would have more information on the subject. | AI Event: AI computed an answer but should hand over to an advisor now |
Transfer to an advisor |
AI - no response (advisor needed) |
The AI does not respond with a sufficiently relevant response to be sent to the customer. | AI Event: AI computed an answer NOT sent, handover to an advisor |
Transfer to an advisor |
Invalid message format returned by AI |
The AI response has an invalid format (image if not supported, too many characters, etc.) | AI Event: Invalid format of message turned by AI |
Transfer to an advisor |
AI - no response (interrupted by user message) |
As of today, we do not support multiple messages sent by the client before the AI can respond to the first one. | AI Event: The user sent another message before the AI could respond |
Transfer to an advisor |
AI API Error |
Our AI partner returns a technical error. This is also the output we take if our AI partner does not respond. | AI Event: AI API error |
Transfer to an advisor |
Exit failure |
The connector encountered a technical problem |
AI Event: Failure exit |
Transfer to an advisor |
(1) Bot events are internal tracking elements of Alcmeon, which allow you to obtain statistics on the use of your bot in the “Analytics / Volumes / Bot Events” section.
4. Measure satisfaction
When the AI Connector step has the output “AI - response (no advisor needed),” the AI response has been sent to the customer, and it is therefore appropriate to ask the customer what they think of this response. To do this, connect the output to an “AI Satisfaction” step.
In this step you can configure:
- The text of the question asking the customer if the AI response seems relevant to him
- The text of the “Yes” response button
- The text of the “No” response button
This stage has 3 exits:
- A “Yes” button to be linked to a message box with a menu to say goodbye to the customer, and possibly to offer them a button to return to the start of the scenario.
- A “No” button to link to the scenario branch to transfer the customer’s message to an advisor
- A “Free Message Entry” output : If the customer doesn’t click on either of the two buttons above but instead types a new message, it’s very likely that this is a continuation of their previous question. We therefore suggest linking this output to the Sub Bot “Capture message content” step earlier in the scenario. Indeed, Alcmeon will indicate to the AI that this new question is a continuation of the previous one to allow the AI to respond with the right context.
5. Manage AI non-response cases
Here is an example illustrating our recommendations from the table above, you can use it as inspiration to build your own scenario!