Your AI chatbot comes with its very own reporting. Access the report at Reporting > AI chatbot.
The AI chatbot reporting contains three sections:
- Owl Intelligence credits: Use this section to review your total available, used, and remaining Owl Intelligence credits for the month. Running out of credits will prevent your chatbot from working.
- Daily usage statistics: Use this section to view daily totals of questions, ratings, and percentages about positive/negative ratings and how many questions the chatbot couldn't answer.
- Questions from readers: Use this section to review the actual questions submitted to the chatbot, whether the chatbot answered them, the articles used to answer it, and more. You can either use this report standalone or use it as a drilldown for the days that seem especially interesting in the Daily usage statistics.
Use the links below to jump to more information on each section.
The Owl Intelligence credits section shows you your total number of available monthly Owl Intelligence credits, how many credits you've used, and how many are remaining.
To access this information, go to Reporting > AI chatbot. Owl intelligence credits is the first section on the page. Your credits will display in a format like this:
This month's credit usage
57 of 1000 credits used (943 remaining)
A credit is used when a reader's question returns at least one article from your knowledge base, regardless of whether the AI is ultimately able to generate an answer from those articles. If the search returns no articles, the AI is never called and no credit is used. In practice, the vast majority of questions return at least one article, so you should expect nearly all questions to use a credit.
Are you running low on AI credits for the month? Follow the instructions in How can I buy extra AI credits? to buy some additional credits without changing your existing plan subscription.
Go to Reporting > AI chatbot to view your AI chatbot's Daily usage statistics.
Use the controls at the top to display the data for the last:
- 7 days
- 30 days
- 90 days
- 180 days
The Daily usage statistics will show the following in a reporting dashboard:
- Total questions asked
- Average answer rate (the percentage of questions the chatbot could answer)
- Average satisfaction rate (the percentage of positive ratings)
- A chart displaying the question volume over time
- A chart displaying the answer rate over time
- A statistics table showing you how much the chatbot is being used each day, including:
- Total Questions: The total number of questions submitted to the chatbot for the day.
- Total Ratings: The number of readers who submitted a rating after receiving an answer to their question.
- % Positive: Of the Total Ratings, the percent who rated the answer with a thumbs up.
- % Negative: Of the Total Ratings, the percent who rated the answer with a thumbs down.
- % Unrated: Of the Total Ratings, the percent who didn't provide a rating.
- % Couldn't Answer: The percent of the Total Questions that the AI chatbot was unable to answer.
- Date: The date the statistics apply to.
To drill into any days, use the Questions from readers report further down on the page.
The Questions from readers section of the AI chatbot reporting lets you review exactly what your chatbot readers have been searching for.
To access it, go to Reporting > AI chatbot.
In the Questions from readers section, use the Today, This week, or This month control to pick a preset time period or select the Select date... field to enter a specific date. to select the date(s) you want to review questions for. The table below updates based on your selection.
Then you can review:
- Question: The question a reader entered into the chatbot.
- Answered: Indicates whether the question was answered (green checkmark icon) or not (red x icon).
- Rating: Indicates whether the reader gave the response a thumbs-up or thumbs-down rating.
- Reader's reader groups: If your knowledge base uses reader groups, this column lists which groups the reader who asked the question belonged to. This can help you determine if the reason a question wasn't answered is because the reader didn't have access to the content that would have answered it.
- Articles used to answer: The sources the chatbot used to answer the question. Use these sources to determine if the chatbot used the correct content to answer the question. It can help you dial in words within the content to surface the correct content in the future.
- Submitted by: The name and email of any logged in reader who submitted the question.
- Time submitted: The time the question was submitted, in the time zone for the knowledge base.
Select Export CSV to download the current data in CSV format.