Article Recommendations

Surface relevant knowledge base articles for every ticket. AI matches ticket content to your help articles, helping you resolve issues faster.

What Are Article Recommendations?

Article recommendations use semantic search to find knowledge base articles that match the ticket's content and context—not just keyword matches, but meaning-based matches.

Example

Ticket: "How do I get my money back? Product doesn't work."

Recommended articles:

1. "Returns and Refunds Policy" (92% match)
2. "How to Request a Refund" (88% match)
3. "Product Troubleshooting Guide" (75% match)

How It Works

Semantic Search

Unlike keyword search, semantic search understands meaning:

Vector Embeddings

1. Your articles are converted to mathematical vectors
2. Ticket content is converted to a vector
3. AI finds articles with similar vectors
4. Results ranked by similarity

Where to See Recommendations

AI Sidebar

Open any ticket to see "Related Articles" in the AI Insights panel:

📚 Related Articles
• Returns and Refunds (92%)
  → Share with customer
• Troubleshooting Guide (75%)
  → Reference for resolution

Inline Suggestions

While composing replies, article suggestions may appear contextually.

Using Recommendations

Share with Customer

Found a helpful article? Share it:

1. Click the article in recommendations
2. Click "Share with customer"
3. Link is inserted into your reply

Reference for Resolution

Use articles for your own reference:

1. Click to open article
2. Read solution steps
3. Apply to customer's situation
4. Paraphrase in your reply

Quick Preview

Hover over an article to see:

• Title and excerpt
• Last updated date
• View count
• Helpfulness rating

Match Quality

Confidence Scores

Factors Affecting Matches

• Article content: Comprehensive articles match better
• Ticket clarity: Clear tickets get better matches
• KB coverage: More articles = more potential matches
• Article freshness: Updated articles may rank higher

Improving Recommendations

Build a Better Knowledge Base

More, better articles = better recommendations:

1. Cover common issues: FAQs, how-tos
2. Use clear language: Match how customers speak
3. Keep updated: Remove outdated content
4. Add variations: Different ways to ask the same thing

Article Optimization Tips

Good article titles:

• "How to Request a Refund"
• "Troubleshooting Battery Issues"
• "Setting Up Your Account"

Avoid:

• "FAQ"
• "Info"
• "Read This"

Include synonyms:

"Returns (Refunds, Money Back, Send Back)"

Recommendation Settings

Configure in Settings → AI

• Minimum confidence: Only show above X% (default: 60%)
• Maximum results: How many to show (default: 3)
• Article collections: Which collections to search
• Include drafts: Whether to include unpublished articles

No Articles Found?

Why It Happens

• Missing KB content: No articles cover this topic
• Low match confidence: Articles exist but don't match well
• New issue type: Not yet documented

What to Do

1. Answer the customer: Don't wait for KB
2. Note the gap: This is a content opportunity
3. Create the article: Document the solution
4. Future tickets: Will now have recommendations

Linking Articles in Replies

Insert Article Link

1. View recommended article
2. Click "Insert link"
3. Link appears in your reply composer

Best Practices

When sharing articles:

Good:

"I've found an article that walks through this step-by-step: [How to Request a Refund]. Let me know if you have questions after reviewing it!"

Avoid:

"See this article: [link]" (Too brief, feels dismissive)

Analytics

Article Performance

Track which articles are most recommended and used:

• Recommendation frequency: How often suggested
• Share rate: How often shared with customers
• Resolution correlation: Does sharing help resolution?

Finding Gaps

Identify needed content:

• Tickets with no recommendations
• Low-confidence matches
• Frequently asked questions without articles

Troubleshooting

"No articles recommended"

• KB may not have relevant content
• Ticket may be too short/vague
• Check KB has published articles

"Wrong articles recommended"

• Article content may not match title
• Update article with relevant keywords
• AI learns from usage patterns

"Articles not updating"

• New/edited articles need reindexing
• May take up to 1 hour
• Force reindex in Settings → KB

For Your Knowledge Base

Article recommendations work best when your KB is:

1. Comprehensive: Cover common issues
2. Well-written: Clear, customer-friendly language
3. Current: Updated information
4. Organized: Logical collections
5. Findable: Good titles and content

See Creating Articles for KB best practices.

← Tag Suggestions | Similar Tickets →