Setup
Can I turn off all recommendations?
Yes, you can turn off recommendations for your account. To do this:
- Access Conversation Assist.
- Click Settings from the menu at the top.
- Under General, click Suspend.
Can I use the same agent group or profile in multiple add-ons within a rule?
Yes. But keep in mind that the order of the add-ons matters: At runtime, the add-ons are evaluated in order, and the first one that’s matched is executed.
Agent experience
Why aren’t my knowledge base answers formatted?
Currently, only hyperlinks are supported.
When I join a bot a conversation, the bot doesn’t always begin the flow that I expect. Why is this?
When you delegate a bot to a conversation, the specific bot flow that is triggered depends on the consumer’s most recent message. This is the case for bots added to conversations using inline recommendations and also for bots added using the On-Demand Recommendations widget.
How are recommendations made?
Step 1: Compile a candidate list of answers
To begin with, the consumer's query is evaluated and classified. If the query only offers information, then no recommended answers are gathered. Here are examples of consumer queries that only offer info:
- My name is Jane Doe.
- My account number is abc1234567.
- I bought the item last month, and the order number is GA2342345.
- My address is 123 4th Avenue, Apt. 2A, New York, New York 10010.
Keep in mind that a consumer query might offer information and also contain an intent, such as, “My account number is abc1234567. Can you tell me my balance?” In these cases, recommended answers are still gathered.
In all other cases (intentful message, small talk, etc.), recommended answers are gathered.
The classification behavior discussed above is in Beta release. We've only rolled this out to brands with accounts in the APAC region so far. We're working to gain learnings from this limited release before rolling it out to all regions.
Next, the system then gathers recommended answers according to the rules specified for knowledge bases. All those found are added to a list of candidates.
Curious about the knowledge base search? Under the hood, Conversation Assist automatically searches a knowledge base for articles (answers) using the KnowledgeAI search offering.
Step 2: Compile a candidate list of bots
Bots work a little differently from answers because bots are recommended regardless of the type of consumer query.
The system gathers recommended bots according to the rules specified for bots. All those found are added to a list of candidates.
Step 3: Determine what to recommended
The system evaluates the candidate lists of answers and bots and chooses those ranked highest by relevance score. The rules for this are as follows:
-
Which recommendations are included? First, include all answer (article) recommendations. Second, include all bot recommendations. This means that answers are included before bots even when the answer scores are lower than that of the top bot recommendation.
-
How are the recommendations ordered? Within each subgroup of recommendations (answers, bots), sort the recommendations by relevance score in descending order so that the higher the score, the higher the recommendation.
Example 1
Assume the system has compiled the following final candidate list, sorted below by confidence score:
- Answer (article) in knowledge base 1 = 90%
- Answer (article) in knowledge base 2 = 60%
- Bot 1 = 100%
- Bot 2 = 80%
If the maximum number of recommendations is set to 2, then only the answers are recommended to the agent. The agent doesn’t see any bot recommendations because answer recommendations are always included first. But if maximum number of recommendations is set to 3, then the two answers and the top-scoring bot are recommended.
Example 2
Assume the system has compiled the following final candidate list, sorted below by confidence socre:
- Answer (article) in knowledge base 3 = 100%
- Answer (article) in knowledge base 2 = 80%
- Answer (article) in knowledge base 4 = 75%
- Answer (article) in knowledge base 1 = 72%
- Bot 1 = 100%
- Bot 2 = 75%
If the maximum number of recommendations is set to 4, then only the answers are recommended to the agent. Here again, the agent doesn’t see any bot recommendations because answer recommendations are always included first. But if the maximum number of recommendations is set to 5, then all of the answers and the top-scoring bot are recommended.
Agent feedback (thumbs up, thumbs down)
Is the feedback of my agents captured in the reports that are available for download?
Yes! Learn about agent feedback reports.
Agents can provide feedback on bot/answer recommendations by clicking “thumbs up” or “thumbs down” on the recommendation. They can also edit recommended answers and send the edited answers to consumers. Do these actions influence what recommendations are subsequently offered?
No, not at this time.
In the case of recommended answers, does editing the answer before sending it, or clicking “thumbs up” or “thumbs down” on the recommended answer influence the knowledge base in KnowledgeAI?
No, not at this time.
Is agent feedback used to enhance the underlying recommendation engine?
No, not at this time.
Metrics and reporting
On the Home page, in the Recommendations widget, the Summary data does not align with the graph data. Why is this?
There’s a bug in the Summary information in this widget. It should show the number of used recommendations “over” the number of offered recommendations, where the latter (the denominator) reflects the recommendations that were and weren’t used.
Currently, the denominator is incorrect. It reflects only the recommendations that weren’t used.