What is AI Recommendations on Intelligent Analytics?
If you still have questions or prefer to get help directly from an agent, please submit a request.
We’ll get back to you as soon as possible.
- Getting Started
- Bot Building
- Conversation Design
Code Step Integration Static Step Integration Shopify Integration SETU Integration Exotel Integration CIBIL integration Freshdesk KMS Integration PayU Integration Zendesk Guide Integration Twilio Integration Razorpay Integration LeadSquared Integration USU(Unymira) Integration Helo(VivaConnect) Integration Salesforce KMS Integration Stripe Integration PayPal Integration CleverTap Integration Fynd Integration HubSpot Integration Magento Integration WooCommerce Integration
- Agent Setup
- Analytics & Reporting
- Proactive Messaging
- Commerce Plus
- Troubleshooting Guides
- Release Notes
Table of ContentsIntroductionHow to use AI Recommendations?User Messages recommendations for Existing StepsSteps that already exist on your botRecommendationsUser Messages recommendations to Create New StepsRecommendations to create new stepsRecommendationsWhy use AI Recommendations?FAQs
Intelligent Analytics is the tool that provides information about how your IVA is performing. It also provides feedback on your IVA and points out the problem areas. Using these learnings, you can improve conversations & interactions that users have with the IVA and take better decisions to build a superior user experience.
AI Recommendations is one such feature on Analytics that identifies issues with your IVA regarding missing user intents and training deficit on existing intents. It uses an algorithm that intelligently analyses user messages which the bot was not able to understand. Instead of manually analyzing 1,000s of interactions every day to find out those problem areas, we optimize this process with AI.
The AI Recommendations algorithm provides you all the key insights to understand user interactions to train your IVA better and enhance its intelligence further.
So to sum up,
- The algorithm tells you what users are chatting about, which you have not accounted for in the IVA. In other words, missing intents. By adding these intents, the number of IVA breaks reduces over time.
- The algorithm also tells you which intents need more training data to make your IVA more intelligent.
How to use AI Recommendations?
As shown in the above image, you can see two data sets. On the left side, you find intents that require more training data. On the right side, you find missing intents which can be added to the IVA.
To start using AI Recommendations, download the sheet by clicking on the link shown in the below image. You will receive two CSV files.
User Messages recommendations for Existing Steps
Existing steps are intents that are already present on the bot. User messages recommendations for existing steps are phrases that can be added to the ‘user messages’ section of these existing steps.
This will help your bot fetch response to a user’s message correctly from the right step because it will be trained with more utterances of the intent on the same step, thus making the training model stronger.
Steps that already exist on your bot
This is the list of step names existing on the bot. These are already trained with the relevant “user messages” messages and correspond to a user’s intent on Conversation Studio.
Recommendations are user messages that broke your bot or caused smart assist to be triggered. The AI Recommendations algorithm suggests that these should be added to the ‘user messages’ section of existing steps on your bot. The volume of the similar intents is also shown in this section. In order to see all the recommendations, you can click on the link shown above to get the detailed report over email.
The downloaded CSV contains 5 columns -
- Steps that already exist on your bot
- User Messages Recommendations
- Closest User Messages present on the step - These are the User Messages already existing on the step.
- Link to Open and Add Existing Step - This is the link of the step so that you can directly navigate to the required step to add the User Messages.
- Confidence Score - This is the confidence score of AI Recommendations algorithms of matching the recommended user messages to the existing steps.
Here is a snippet of the CSV
User Messages recommendations to Create New Steps
The Create New Steps section helps you identify intents that are out-of-scope of the bot as well as identify what are the new user utterances surfacing over the bot.
These messages will help you identify and understand what should be the new capabilities that can be added to the bot.
Recommendations to create new steps
The AI Recommendations algorithm has identified similar intents and clustered them together to understand how we can make them into new steps. These recommendations can be added as new steps to increase the intelligence and capability of your bot to reply to user's queries.
You can see the volume of the similar new intents in this section. In order to see all the recommendations, you can click on the link shown above to get the detailed report over email.
The downloaded CSV contains four columns -
- Recommendations to create new steps - These are the recommended names of new steps that can be created with the corresponding user messages in the next column.
- User Messages Recommendations - These are the user messages to which the IVA responded with Smart Assist and Bot Break Messages. Then can be used as User Messages on the new steps.
- Cluster ID - This helps you understand the count of clusters formed. AI algorithm has identified similar intents and clustered them together to understand how we can make them into new steps.
- Confidence Score of the Cluster Identified - This column will help you understand the similarity in the messages to cluster them together.
Here is a snippet of the CSV -
Why use AI Recommendations?
AI Recommendations will not only reduce the manual effort required to go through every single message to identify IVA breaks but also lead to algorithmically improving the training data based on statistically analyzed user data.
Using AI Recommendations you can -
- Identify the Out of Scope queries and find new user intents that can be added to the IVA in order to improve the user experience. This can reduce IVA maintenance by 30%.
- Categorize similar user utterances to a single bucket of intent that will lead to fewer bot breaks, and in turn increase the Smart Funnel completion rate of in-scope intent.
How to enable AI Recommendations?
In order to enable AI Recommendations, you will have to connect with our Customer Success team at email@example.com.
When will the data get populated?
If the bot went live today then all the messages which have been sent to the bot will be processed and analyzed to create User Messages Recommendations. You will be able to check this data from the next day onwards.