Generate recommendations

Once you have finalised your review, you can generate recommended updates to your NLP Question training phrases. Generating and applying recommendations happens in three stages:

1. When you request recommendations, the inGenious AI platform suggests up to five NLP questions that could be improved with changes to training phrases.
   You can select which questions you want to generate recommended changes for.
2. The inGenious AI platform suggests up to five training phrases to add or remove for each question you selected.
3. You select which recommendations to apply. The inGenious AI platform will apply the changes to the questions for you. You can review and publish the changes as normal.
   We recommend using versions to keep your regular content edits separate from recommended changes.

Once you have applied recommended changes from a review, you can't generate and apply new recommendations from the same review. You will need to complete another review.

Generating suggested questions

To generate suggested questions:

1. Open the team and chatbot you want to generate recommendations for.
2. In the left-navigation, click Improve.
3. Click Reviews.
4. Open the finalised review you want to use to generate recommendations.
   Each review can only be used to generate recommendations once.
5. Click the Summary tab.
6. Click Generate recommendations.

The inGenious AI platform will suggest up to five NLP questions to improve. This will take up to ten minutes. You will receive an email when the suggested questions are ready.

Generating recommended training phrase updates

When the suggested questions are ready, the person who requested them will receive an email with a link to the review summary. If that wasn't you, you can still generate recommended training phrases for the questions in the Review summary page:

1. Open the team and chatbot you want to generate recommendations for.
2. In the left-navigation, click Improve.
3. Click Reviews.
4. Open the finalised review you want to generate recommendations from.
5. Click the Summary tab.
   The questions are displayed in the Summary tab with the conversation and passage they're defined in.
6. By default, all suggested questions are selected. If you want to limit the recommendations to only some of the questions, click the padlock icon in the top left of the questions list and select the questions you want to use.
   We recommend generating recommendations for all five questions. You can select which recommendations you want to apply later.
7. Click Generate recommendations.

inGenious AI will generate recommended updates to your training phrases. This may take up to two hours for larger chatbots. You will receive an email when the recommendations are ready.

Applying recommendations

When the recommendations are ready, the person who requested them will receive an email with a link to the review summary. If that wasn't you, you can still review and apply the changes in the Review summary page:

1. Open the team and chatbot you want to generate recommendations for.
2. In the left-navigation, click Improve.
3. Click Reviews.
4. Open the finalised review you want to generate recommendations from.
5. Click the Summary tab.
   The recommended changes are displayed under each question.
   Training phrases to be added are shown in green, and training phrases that are to be removed are shown in red.
6. Select the Accept checkbox next to each change you want to apply.
   Select all the changes you want to apply. Once you have applied your selected recommendations, you will not be able to use this review to automatically apply other changes.
7. Click Apply recommendations.
   The changes you have selected are applied to the questions.
8. Click Review and Publish to go to the Ready to publish content screen to publish your changes.

When recommended changes are automatically applied to questions that are in published status, those questions are moved directly to ready to publish. Questions that were in draft or ready to review will stay in their current review status, and will not be moved to ready to publish. We recommend using versions to keep your regular content edits separate from recommended changes.