Overview
Autopilot is Canny's AI-powered suite of tools to automate the process of collecting, organizing, and refining feedback.
Connect a source to have Canny automatically extract feedback from a tool your team is using.
Automate merging duplicate posts and dealing with spam.
Improve communication using Smart Replies.
Summarize user input on an idea using Comment Summaries.
The Benefits:
Get more feedback with automated detection of feature requests.
Organize feedback with an inbox that surfaces new items with quick actions.
Identify spam with automated spam detection.
Automate feedback collection a trustworthy tool.
This article covers:
Enabling Autopilot
Manual vs Automated
In the Autopilot area of your Canny settings, you can choose between using Autopilot’s manual or automated mode. The Autopilot view should be available to you in your admin navigation.
Manual
Enabling manual mode will have Autopilot surface suggestions, but your team must go to the Autopilot inbox to approve actions manually. This is what a manual Autopilot inbox looks like:
-
Automated
Enabling automated actions will turn the Autopilot tab into an audit log where you can review actions taken by Canny Autopilot rather than an approval queue where manual actions need to be taken.
Enabling automated actions will allow Canny to automatically:
Create new posts (select which board will receive newly-created posts from the
Autopilot tab in your Canny settings).
Merge duplicate posts.
Enabling automated actions will NOT include:
Autopilot will NOT automatically mark a post as spam.
Posts that are identified as likely spam are surfaced for your team to take action.
See the Things to note section below.
Enabling automation will turn the Autopilot inbox into an audit log of actions taken. Your team will review what Autopilot has already done. Just select the board that Autopilot will use when creating new posts:
Here is what an automated Autopilot inbox looks like:
Here, you can choose whether to instead merge or delete posts that Autopilot has created and review what actions Autopilot has taken.
Autopilot will NOT take automated action to:
Keep a newly-created user post.
When a user creates a post, it will be instantly visible on the board to other users. Autopilot does not keep a user-created post from being visible unless it’s flagged as likely spam.
Mark a post as spam.
Autopilot will mark posts it thinks are likely spam. If the confidence that the post is spam is low, then the post will still be visible to other users. If that confidence is moderate or high, the post will not be visible unless your team approves the post from the Autopilot inbox.
Sources
A "source" is a connected integration that Canny Autopilot monitors to extract potential feedback opportunities, identify likely spam, and automatically merge duplicate posts.
Autopilot can extract insights from several tools.
These sources will count toward your plan's integration limit:
These tools can be connected as sources but do NOT count toward your plan's integration limit:
Trustpilot
Apple App Store
Google Play Store
G2
Capterra
TrustRadius
ProductHunt
Salesforce AppExchange
WordPress
Shopify
Zapier (only available on the Growth plan)
The Canny API's AI endpoint
Connected sources are listed in the left-hand column:
Things to note about sources:
PII (Personal Identifying Information) fields are not processed by AI.
Conversations are only processed after they are closed out in Intercom, Zendesk, etc.
When first connecting a customer communications integration (Intercom, Zendesk Help Scout, etc.), Canny will process closed conversations from the previous 30 days.
When connecting a review integration (App Store, Google Play, Capterra, etc.), Canny will process reviews from the previous 60 days.
New conversations will be processed as older suggestions are resolved. At 100 pending items, Autopilot processing will be paused until actions are taken to clear the pending suggestions in the Autopilot inbox.
In Intercom only: If a support rep manually tracks feedback, Canny will skip processing that conversation.
Credits
1 Autopilot Credit = $0.10 (USD)
Every Autopilot-enabled action uses credits. All plans include a set number of credits to use each month:
Free: 25
Starter: 100
Growth: 500
Business: 1,000+ (contact us for Business plan pricing)
Additional credits
Once your plan's allotment of credits (see above) have all been used, you can manually ensure Autopilot continues processing conversations using either or both of the following options:
Set a monthly credit cap beyond your plan’s included credits.
This means credits will be purchased in groups of 100 ($10) until the limit is hit.
Just enable the toggle on the billing page and set a hard limit:
Select a top-up amount
Instead of $10 at a time, your team purchases a bulk allotment of additional credits:
When is a credit used?
A newly-created post is checked for duplicates Canny: 1 credit
If the ‘Canny Portal’ toggle is enabled in your Autopilot settings, all new posts created in Canny will be compared to existing posts to ensure no duplicates.
If a new post is similar to an existing post, Autopilot can merge the newly created post as a vote and add that user's input as an internal comment to the existing post.
A text conversation, app review, or API call is processed by Autopilot (regardless of whether it results in a suggestion): 1 credit
This applies to sources like Intercom, Help Scout, App Store, and the API.
So if your team has 100 Zendesk cases in a month, that will use 100 credits, even if only 50 suggestions were generated after Autopilot processed them.
An audio call is processed by Autopilot: 5 credits
A credit is charged regardless of whether the conversation results in a suggestion.
So if you have 100 Gong calls in a month, that will use 500 credits, even if only 50 suggestions were generated after Autopilot processed them.
A Smart Reply is generated: 1 credit
A Comment Summary is generated: 1 credit
Note: Spam identification does NOT use up credits
Credit billing
Limits on additional credits can be set to prevent overages
Credits do not roll over month-to-month. The limit is reset with each billing cycle.
A credit card is required to be on file for additional credits.
Discounts do apply to credits. The only exception is the early-stage startup discount. The early-stage discount will not apply to your Autopilot credits.
Manual operations
Manual operations:
1. Review suggestions to create a new post from a source
Autopilot will look through your sources (e.g. app store reviews, sales calls, support conversations with clients, etc.) and extract relevant suggestions to add feedback.
Actions
Thumbs up (👍): Create the post
NOTE: Creating a post does not notify the end-user in any way
Thumbs down (👎): Do not create the post. Take no action and dismiss this suggestion
Add vote to other post (···): Add this as a vote (with the text as an internal comment) to an existing post
2. Review newly-created posts
When a new post is created in Canny by an end-user, it will automatically appear in Autopilot with Canny as the source. Posts created by fellow Canny admins will NOT be surfaced in the Autopilot tab.
This way, your team can easily approve, delete, or merge newly-created posts. To be clear, this will not prevent the post from appearing on the board instantly when the user creates it. This is only for review purposes.
Actions
Thumbs up (👍): Keeps the post on the board (Again, the post already exists).
Thumbs down (👎): Delete this post from the board.
Add vote to other post (···): Add this as a vote (with the text as an internal comment) to an existing post.
3. Merge duplicate posts
Autopilot will automatically identify posts that are likely duplicates and recommend merging them. This can be automated so that Canny will simply merge new duplicates into existing posts. Then, you can review the Autopilot page as an audit log rather than a queue awaiting your action.
Actions
Thumbs up (👍): Merge the new duplicate into the older existing post.
Thumbs down (👎): Dismisses the suggestion with no action. Both posts will stay on the board as separate posts. ****Nothing is deleted.
Add vote to other post (···): Merge the newly-created post to a different existing post.
4. Spam identification (does not use credits)
Canny will analyze posts for potential spam. The "Confidence" column indicates how confident Canny is that this is spam. The higher the confidence, the more sure the system is that this is spam.
Actions:
Mark as spam:
Delete the post
Delete all other posts, votes, or comments by that user
Delete the user's profile from Canny
Block that IP address from posting again
Not spam: Leave the post on the board
Delete post: Simply delete that post without taking any further action on that user
Filters
Limit the view of the Autopilot page as needed. If there are multiple people processing the Autopilot inbox, it will likely make sense to use filters.
The Autopilot API endpoint
An Autopilot endpoint is available on the Canny API here if you'd like to programmatically send feedback to Canny for processing. Each call to the Autopilot endpoint will use 1 credit.
Use the AI endpoint to extract feedback from places like:
App store reviews
Social media
Emails
Call transcripts
... and more!
Canny’s Zapier integration can also send items to Canny Autopilot to be scanned. Just a reminder that the Zapier integration is only available on the Growth plan.
Things to note
Canny will only sync data from conversations where the most recent interaction occurred within the last 30 days.
Canny only syncs new conversations if the inbox has less than 100 pending items. So as you go through the inbox and clear pending items, new suggestions will keep coming in.
If a post is marked as closed or complete, Canny will automatically dismiss any pending Autopilot suggestions related to them.
Discounts do apply to credits. The only exception is the early-stage startup discount. That discount will not apply to your Autopilot credits.
Autopilot is only visible to admins with the Owner or Manager admin role.
Identifying spam posts does NOT use credits.
An AI endpoint is available on the Canny API here if you'd like to programmatically send feedback to Canny for processing.
Autopilot processes conversations from newest to oldest.
By default, only low-confidence spam posts are shown on your board without approval in Autopilot. Spam posts marked as moderate or high confidence will not be shown on the board until approved in Autopilot.
If you find that a post is missing that you expect to appear, look at the “high confidence” posts to make sure nothing is inaccurately flagged as spam by Autopilot.
When marking an action as spam, the action will mark the user as a spam user: • Delete all of their posts, comments, and votes. • Ban them from voting, along with creating future posts and comments.
Related Articles:
Keywords: auto, auto-pilot