Summary
Suggestions helps you improve your AI Editor by learning from your team's past customer conversations. Each week, it analyzes a sample of highly rated and poorly rated conversations and suggests updates to your Knowledge Base and AI Editor's communication style. You can review, edit, approve, or skip each suggestion, so you stay in control of every change.
Note: AI Editor Suggestions are available only on Wati Pro and Business plans.
Instructions
Who can use Suggestions?
Suggestions is available to account admins who manage the Knowledge Base and AI Editor settings.
Your account must meet the following requirements:
Your account must be at least 90 days old.
Your account must have at least 50 quality-rated conversations.
If your account does not meet these requirements, the Suggestions tab displays a message while your conversation history builds up.
How Suggestions work
Suggestions analyzes a sample of your recent conversations each week, focusing on conversations with the highest and lowest ratings. Based on these conversations, it generates suggestions that can help improve your AI Editor.
Suggestions can generate up to 12 suggestions per week across 4 categories:
Suggestion type | Based on | What it suggests |
Q&A Pair | Highly rated conversations | A customer question and a strong answer from your team that you can add to your Knowledge Base. |
Positive Pattern | Highly rated conversations | A communication approach that worked well, such as acknowledging a customer's frustration before explaining the next steps. |
Anti-Pattern | Poorly rated conversations | A communication habit that may have frustrated customers and that your AI Editor should avoid. |
Knowledge Gap | Poorly rated conversations | A topic where your team struggled to provide a confident answer, indicating that you may need to add more information to your Knowledge Base. |
Nothing is changed automatically. Every suggestion requires your approval before it is applied.
Find Suggestions
Go to your Knowledge Base.
Open the Suggestions tab.
The Suggestions tab appears alongside the Q&A and Documents tabs.
An orange dot appears on the Suggestions tab when you have new suggestions to review.
Review suggestions
Each suggestion appears as a card with a confidence label. You can choose one of the following actions:
Approve as-is: Apply the suggestion without making changes.
Edit, then approve: Modify the suggestion before applying it.
Skip: Dismiss the suggestion without applying it.
Skipped suggestions won't appear again for 60 days.
If you don't take action, the suggestion expires after 60 days. Suggestions are never approved automatically.
Where approved suggestions are applied
The destination of an approved suggestion depends on its type:
Q&A Pairs: Added to the Q&A tab in your Knowledge Base. Your AI Editor can use this information when answering future questions.
Positive Patterns: Added to the Writing Style & Behavior section of your AI Editor settings. Your AI Agent uses this guidance when communicating with customers.
Anti-Patterns: Added to the Behaviors to Avoid list in Writing Style & Behavior. Your AI Editor uses this guidance to avoid repeating the identified behavior.
Knowledge Gaps: Not stored automatically. Use these suggestions to identify information that should be added to your Knowledge Base.
Manage Writing Style & Behavior
The existing Brand Tone setting has been expanded into Writing Style & Behavior.
This section contains approved communication guidance from Suggestions as well as guidelines that you add manually. Positive patterns and anti-patterns approved through Suggestions are added here automatically.
Sort and filter suggestions
Use Sort by and Filters to quickly find the suggestions you want to review.
Sort suggestions
Use Sort by to change the order in which suggestions appear. You can choose from:
High to Low Confidence: Shows suggestions with the highest confidence first.
Recent First: Shows the most recently generated suggestions first.
Filter suggestions
Use Filters to show only specific types of suggestions. The available filters are:
Q&A Pairs
Communication Pattern
Pattern to Avoid
Knowledge Gap
Each filter is a checkbox. Select one or more filters to show suggestions that match your selection. Clear a checkbox to remove that suggestion type from the results.
For example, selecting Communication Pattern and Knowledge Gap shows only suggestions from those 2 categories.
Use the available controls to narrow the list, with High to Low Confidence or Recent First sorting and checkbox filters for each suggestion type so the most relevant recommendations appear first.
Suggestions limits and safeguards
Suggestions includes several safeguards to keep recommendations relevant and manageable:
Up to 12 suggestions per week: Suggestions generates a maximum of 5 Q&A pairs, 2 positive patterns, 2 anti-patterns, and 3 knowledge gaps each week.
Confidence labels: Each suggestion is labeled High, Medium, or Low confidence to help you assess it before approving it.
Duplicate prevention: Suggestions that are very similar to existing Knowledge Base content or recently dismissed suggestions are not shown again.
Privacy protection: Suggestions do not include personal information such as customer names, order IDs, phone numbers, or addresses.
60-day expiration: Suggestions that are skipped or left unreviewed won't appear again for 60 days.
Optional feature: If Suggestions is disabled for your account, the Suggestions tab does not appear and no suggestions are generated.
For new accounts
If your account is less than 90 days old or does not yet have enough quality-rated conversations, the Suggestions tab displays:
Building your conversation history...
Once your account meets the required thresholds, Suggestions can begin generating recommendations from your conversations.
Frequently Asked Questions (FAQs)
Suggestions Overview
1. What are AI Editor Suggestions?
AI Editor Suggestions analyzes a sample of highly rated and poorly rated customer conversations each week and recommends improvements to your Knowledge Base and AI Editor's communication style. Suggestions are reviewed and approved manually before any changes are applied.
2. Who can use AI Editor Suggestions?
AI Editor Suggestions is available to account admins who manage Knowledge Base and AI Editor settings on the Pro and Business plans. The account must also be at least 90 days old and have at least 50 quality-rated conversations.
Suggestion Types and Review
3. What types of suggestions can AI Editor Suggestions generate?
AI Editor Suggestions can generate 4 types of recommendations:
Q&A Pair: A customer question and a strong answer from a highly rated conversation that can be added to the Knowledge Base.
Positive Pattern: A communication approach that worked well in a highly rated conversation.
Anti-Pattern: A communication habit from a poorly rated conversation that the AI Editor should avoid.
Knowledge Gap: A topic where the team struggled to provide a confident answer and may need to add information to the Knowledge Base.
4. How many suggestions can AI Editor Suggestions generate each week?
AI Editor Suggestions can generate up to 12 suggestions per week: 5 Q&A pairs, 2 positive patterns, 2 anti-patterns, and 3 knowledge gaps.
5. What can I do with a suggestion?
You can Approve as-is, Edit, then approve, or Skip a suggestion. Suggestions are never applied automatically. Skipped or unreviewed suggestions expire after 60 days.
Suggestion Management
6. How can I sort and filter suggestions?
Use Sort by to arrange suggestions by High to Low Confidence or Recent First. Use Filters to show specific suggestion types: Q&A Pairs, Communication Pattern, Pattern to Avoid, or Knowledge Gap. You can select multiple filters at the same time.
7. Where are approved suggestions applied?
Approved Q&A Pairs are added to the Q&A tab in the Knowledge Base. Approved Positive Patterns are added to Writing Style & Behavior, while approved Anti-Patterns are added to Behaviors to Avoid. Knowledge Gaps are not stored automatically.
Suggestion Safeguards
8. How does AI Editor Suggestions prevent irrelevant or sensitive recommendations?
Suggestions that are very similar to existing Knowledge Base content or recently dismissed suggestions are not shown again. Suggestions also do not include personal information such as customer names, order IDs, phone numbers, or addresses. Each suggestion is labeled High, Medium, or Low confidence.