Coming soon for Agent+ customers
See how AI, tagging, and saved content are performing, so you can speed up replies, fix broken workflows, and prioritize what to improve next.
1. Start With the Two Most Important KPIs
🧠 Agent+ Involvement
What it shows:
How many replies were completed using any Agent+ feature—AI replies, saved replies, batch mode, reply workflows, you name it.
What to do with it:
Use this to spot adoption. High numbers mean your team is using tools that help clear queues faster. Low numbers? Time to coach the team on using batch mode, prompts, or workflows more effectively.
⚙️ Agent+ Assist Rate
What it shows:
The % of all replies that used at least one Agent+ feature.
What to do with it:
This rate should be high. If it’s low, it means your AI tools aren’t being used enough—and that’s a drag on speed. Help the team by:
Reviewing workflows
Recommending batch mode
Creating shortcuts to high-volume prompts and replies
2. Monitor Reply Quality with Approve, Edit, and Reject Rates
When agents review AI replies, they can approve, edit, or reject. These three metrics tell you how that’s going.
✅ Approve Rate
Goal: Keep it high
If it’s low, something’s off—usually with the prompt, the knowledge, or the tone.
→ Action: Go back to your [Optimize Loop]
Refine prompt rules (add Do’s & Don’ts)
Add or update Knowledge Base content
Create a custom snippet that gets to the point faster
✏️ Edit Rate
Goal: Medium is healthy
A little editing is fine. Rising edits? That’s a sign the prompt or rule might need tightening.
→ Action: Check if rules are too vague or if knowledge is missing detail.
❌ Reject Rate
Goal: Keep it low
High rejection means the AI is missing the mark.
→ Action: Prioritize fixes. This is a clear signal something’s broken—usually the inputs.
3. Tagging: Use Conversation Tags to Route and Respond Faster
Agent+ tags each message it thinks deserves a reply. You’ll see:
Total tagged as engageable
Top tags (FAQ, Complaint, Feedback, etc.)
What to do with this:
Use tag spikes to decide where to build new workflows
If "Buying Intent" is surging, create a dedicated prompt or reply set
Think of tags as swim lanes—each should have a response system behind it
4. Knowledge: Build Smarter Replies Over Time
This section is your feedback loop for how well your content is helping AI write good replies.
🔍 Knowledge Used
How many replies used at least one linked knowledge source (like a help page, article, or doc)
→ Action: High = good. Low = check if articles are too hard to find, too vague, or not linked to the right prompts.
📉 Knowledge Gaps
Where AI wanted to use a source, but nothing was found.
→ Action: Treat this like your weekly to-do list
Add missing content (FAQs, help docs, snippets)
Connect the right sources to prompts or workflows
Add fallback rules in prompts to guide the model better
5. Knowledge Gaps Feed: Real Messages That Had No Help
This feed shows actual messages where the AI couldn’t find a knowledge source to reference.
What to do with it:
Review this weekly—it’s a goldmine for content fixes
Spot recurring questions and add short answers or full pages
If an answer already exists, improve tagging or linking so the AI can find it
Track before/after changes by watching approve and reject rates
📌 This is where long-term quality lives. Fix gaps here, and the whole system gets smarter.
6. Engage Library Usage: Refine What Your Team Is Actually Using
Top AI Prompts
Shows which smart prompts are getting used the most in replies.
→ Action:
If a prompt is barely used, check: is it hard to find? Is it too niche? Are filters wrong?
Remove low performers to reduce clutter
Improve high performers with better rules or fallback content
Top Saved Replies
Your most-used canned replies.
→ Action:
If one reply is doing all the work, create variations so things don’t sound robotic
Update anything outdated or low-performing
Tie replies to tags or workflows to surface them faster
Final Tips: How to Put This Dashboard to Work
Use this dashboard to:
Spot usage issues: Low assist rate = missed speed wins
Catch quality problems: High reject or edit rate = something’s broken
Fix knowledge gaps: Use the feed as your content improvement backlog
Make AI smarter over time: The more accurate your inputs, the better your outputs
Drive measurable impact: Every fix here ties back to agent performance—faster replies, more consistency, and fewer hours spent typing
Pro tip: Keep one eye on this dashboard, and one on your main KPIs—reply time, hours saved, and volume handled. The tighter the loop, the smarter your AI gets.