Direct questions
Last updated
Last updated
Direct questions are a more precise search algorithm where the client asks a direct question, and we can give them a specific answer.
This is a sort of FAQ or standard answers to typical questions. There are often similar questions about the same topic, with the same keywords. By using the "Direct Questions" section, we can set the necessary level of precision, and the bot won't make mistakes understanding which file to use.
This section is more suitable for automation cases:
in sales departments
in HR departments
That is, for those situations where the volume of typical question-answer/bases of knowledge does not exceed 70-80 files.
When using this section, the search goes through file titles, or more specifically, the File Name for Search field:
By default, the file name for search is generated automatically and in English. The English language is used in this case because it is cheaper in terms of token cost, reducing the cost of the dialogue, and the bot understands such names better than in Russian.
If Suvvy does not find a particular file or it doesn't trigger, check that the English version corresponds to the translation, as there are often similar meanings across different languages and the bot may have formed an incorrect heading that doesn't match the actual meaning. In this case, you can correct the English version, or simply write it in Russian (but if you need multilingualism, take into account its application features, more on this here).
Therefore, it's important to create a file with the most concise title, and try to avoid using interrogative sentences in titles.
For example, for queries like "How to get to you?" you need to create a file "Office Address", and Suvvy understands and will call it.
For queries like "How much does it cost?" you need to create a file "Cost", "Price List", or "Client is interested in the cost".
For queries like "How can I pay for your services?" you need to create a file "Payment".
There are different settings within the knowledge base file:
Use - turns the file on/off within the bot's knowledge.
Interrupt Dialog - interrupts Suvvy's responses if this file is triggered.
For example, interruption is needed when you want to transfer the dialogue to a live employee, and so the bot stops giving answers to any customer questions after this response.
Notification in Telegram - allows receiving notifications in Telegram when the specified file triggers:
To connect notifications, besides enabling this tumbler, it is important to subscribe to the bot through a separate section.
The file content represents the information that Suvvy will use when responding. The content of the file, similarly to the system instruction, can be of two types and contain:
a ready-made response for the client
an instruction for Suvvy
an instruction and a response simultaneously
Ready answer for the client - implies we give a ready-made reply to the client , which Suvvy will take and use as an answer, adapting to the specific conversation context.
Instruction for Suvvy - here we can indicate not a specific answer, but the direction of what Suvvy should say without a precise wording.
Instruction and response simultaneously - is a hybrid of the first and second options, where there can be both an instruction and specific answers within the knowledge base file.
Send Follow-Up - this is the ability to send additional messages to the client if they did not reply to the last message.
Sending additional messages is possible both with reference to knowledge base files and randomly through settings in additional settings.
Additional messages can be sent in various ways in terms of sending time:
Sending after a specified number of minutes
Sending at a specific time
If we are talking about the second option, we can send either today at a specific time - in this case, the number of days is indicated as 0, or after a specified number of days - in this case, the necessary number of days and sending time are indicated.
For example, if we look at the screenshot above, the first message is sent after 5 minutes, and the second message (if the client still does not respond) will be sent tomorrow at 9 am.
Regarding the content of the message, there are two options:
Fixed given message
Message generated by Suvvy based on the conversation context and instructions
In the example above, we can see that the first message is a fixed "What do you think", and in the second case, the message will be generated by the model based on the instruction "Remind the client about yourself based on the conversation context".
Image search allows attaching images to a knowledge base file and finding the necessary file for the bot when this image or its part is sent (for example, as a screenshot).
For example, a client saw a photo on social media, takes a screenshot of the screen and sends it to the company with a question "How much does it cost?". When Suvvy receives this image, it first searches for it in knowledge base files, and if it finds it, it checks the file content (which may have a product description, cost, etc.)
The Actions in tab appears when specific CRM channels such as amoCRM, Kommo, Bitrix24, and UseDesk are connected.
This section of the knowledge base file allows performing actions within the CRM when the knowledge base file triggers. For example, when we need to change the deal status in CRM or fill certain fields with specific values.
More details about this section concerning each CRM are described in the Channels section.
The tags section allows marking files with specific features - tags and tracking analytics on which situations in the dialogue trigger most often.
For example, we can mark a file with a target action - in this example - when the client is ready for an online call with an appropriate tag. Then, when this file triggers in the dialogue, the dialogue is marked with this tag. Later, we can review which tags triggered most frequently and in what quantity and percentage for the selected period:
