Automatic Intent Assignment now makes it easier for consumers to indicate their intent. For high-confidence cases, whenever Helpshift AI detects that user text has high confidence matching with the specific intent, then it automatically assigns that intent to the issue which reduces the manual effort of selection by the end user and hence provides a seamless experience.
Prerequisites
Enablement steps
On the Helpshift toolbar, click Settings.
Then, in the navigation pane, scroll to the APP SETTINGS area and click on App for which you want to make changes. The selected app settings page opens.
On the App settings page, click CONFIGURE under the Supported Platforms sections.
The selected platform configuration page opens.Scroll to the User Reply Input Type area for the supported platform, and click one of the intent input types.
Note: Automatic Intent Assignment works for the following two intent input types:Intent Menu + Text
Only Text Input
To enable the Automatic Intent Assignment, turn On the toggle next to it.
Click SAVE & PUBLISH.
For Web, Email, and SMS you can go to below given respective sections and enable/disable Automatic Intent Assignment and update the intent menu:
- Web: Settings → Web Support Portal
Email: Settings → Email Support
SMS: Settings → <respective app page> → SMS → Configure
How does it work?
Automatic Intent Assignment works for the following two intent input types:
Intent Menu + Text
Only Text Input
Note: Automatic Intent Assignment would be done only for High Confidence Intents. No automatic intent assignment would be done for Medium and low Confidence Intents.
About Confidence Score
The intents are scored based on the confidence of the model as HIGH, MEDIUM, or LOW. This confidence tells us how likely a given intent is present in the end user's message.
Automatic intent assignment only considers intents with HIGH confidence, in which case the intent will be assigned to the user's message. If no intent reaches HIGH confidence, automatic intent assignment will not apply the intent.
The confidence score and level mapping is as follows:
HIGH: 50-100% score
MEDIUM: 25-50% score
LOW: 0-25% score
Confidence scores are determined simultaneously for all intents in the model in such a way that the individual scores add up to 100%. As a consequence, only one intent can have 50% or higher confidence. We found that intents with HIGH confidence provide accurate and effective intent assignments.
Please refer to the below-given tables for understanding how Automatic Intent Assignment(AIA) feature works with multiple use cases:
Case 1
Case 2
Case 3
