Your AI voice agents handle calls around the clock, but how well are they doing? The AI Agent Reports page gives you a complete picture of performance—from containment rates to customer sentiment—so you can fine-tune your agents and get more out of them.
To open it, go to Reports in the sidebar and click View Detailed Report on the AI Voice Agent card. The page heading reads AI Voice Agent Reports, and it lives at /reports/ai-agents. It's part of the Analytics suite and needs the AI Agents feature on your plan.
Key Performance Metrics#
When you open the AI Agent Reports page, you'll see six metric cards at the top. Each card shows the current value and compares it to the previous period so you can spot trends at a glance.
Total AI Calls tracks every call your AI agents handled during the selected time period. This number helps you understand your AI's workload and how it changes over time.
Containment Rate is the percentage of calls fully resolved by your AI without human intervention. Higher is better—it means your AI is handling more calls independently. Most businesses aim for 70% or higher.
Self-Service Actions counts calls where your AI reached its goal, like booking an appointment or answering the caller's question. It counts one per call, and transferred calls don't count.
Average Handle Time shows how long calls typically last. Shorter times often mean your AI is efficient, but context matters—complex inquiries naturally take longer.
Escalation Rate is the flip side of containment. It shows what percentage of calls got transferred to a human agent. Lower is better, but some escalation is normal for complex issues.
Cost Savings estimates how much you're saving by having AI handle calls instead of human agents. dialnote calculates this at $3 per contained call, which is the typical cost of a human-handled customer service call.
Quick insight
Look at the trend arrows next to each metric. Green up arrows show improvement, while red down arrows indicate areas that might need attention. Flat lines mean performance stayed consistent.
Understanding Call Outcomes#
Each AI-handled call falls into one of six outcome categories:
- Fully Contained – The AI resolved the call completely. This is your goal for most calls.
- Escalated to Human – The caller was transferred to a team member. Check escalation reasons to understand why.
- Voicemail Left – The AI's call reached a voicemail system.
- Abandoned by Caller – The caller hung up before the AI could help them.
- No Answer – The call failed to connect before anyone answered.
- Call Failed – A technical issue prevented the call from completing.
The calls table shows every AI-handled call with details like caller info, duration, outcome, and sentiment. Click any row to see the full transcript and summary.
Sentiment Analysis#
dialnote analyzes each AI conversation to determine how the caller felt during the interaction:
- 😊 Positive – The caller sounded happy or satisfied
- 😐 Neutral – Neither positive nor negative
- 😟 Negative – The caller seemed frustrated or unhappy
The sentiment breakdown chart shows the distribution across all calls. High positive sentiment usually correlates with good containment rates and effective AI responses.
Improving sentiment
If you see high negative sentiment, review those call recordings. Common causes include AI responses that miss the point, long wait times, or confusing menu options.
Escalation Analysis#
The escalation section shows two things: your containment rate trend over time and your top reasons for escalation.
The Containment Rate Over Time chart plots daily or weekly containment percentages. Look for patterns—did a recent knowledge base update improve performance? Did a new product launch cause more escalations?
The Top Escalation Reasons table lists the top 10 reasons calls got transferred to a human, with a count and each reason's share of all escalations. Calls that ended without a transfer (hang-ups, voicemail) aren't counted here. Common ones you'll see:
- Connected to Team Member – The AI handed the call off and a teammate picked up
- No Team Member Answered – The AI tried to transfer, but nobody answered
- No Team Member Available – No teammate was available to take the call
- Transfer Failed – The handoff didn't go through
Focus on your most common reason first. If "No Team Member Answered" tops the list, check who's in the transfer path and when they're around. If most transfers connect, listen to what callers ask right before the handoff: your AI might need clearer answers. Adjust the agent in AI Agent configuration or expand its knowledge bases.
AI Insights#
At the bottom of the dashboard, the Call Insights section shows automatically generated insights (or "Not enough data to generate insights yet." when nothing triggers). They're split into two categories, with up to three cards each:
Top Opportunities highlights areas for improvement. These trigger on set thresholds—negative sentiment above 20%, escalation rate above 30%, or a containment rate that dropped more than 5% versus the previous period. You might see suggestions like:
- "Containment rate dropped by 8%—review your AI agent's knowledge base"
- "High escalation rate detected: 45% of escalations are due to 'No Team Member Answered'"
- "High negative sentiment detected—review call recordings to find pain points"
What's Working Well celebrates your wins:
- "Containment rate improved by 12% this week"
- "Escalation rate decreased from 35% to 27%"
- "Average handle time reduced by 45 seconds"
Taking action
Each opportunity includes a direct link to act on it. To make the change, open AI Agents in the sidebar, and use the calls table on this page to listen to what went wrong.
Filtering Your Data#
Use the filters at the top of the page to narrow your view:
- Date Range – Pick a preset (Today, Yesterday, Last 7 days, Last 30 days, Last 90 days, Month to date, Quarter to date, Year to date) or set a custom range
- Phone Numbers – Filter to specific phone numbers if you have multiple lines with AI agents
- Tags – Shows up in the filter bar, but it doesn't change the numbers on this report
- Granularity – Switch between Daily and Weekly views for the trend charts
Your filter selections persist in the URL, so you can bookmark specific views or share them with teammates. The calls table has its own Export CSV option if you want the raw rows in a spreadsheet.
What's Next#
- AI Agents overview – set up and configure your voice and message agents
- Sentiment Analysis – how dialnote scores caller sentiment
- Call Scoring – AI call evaluation and quality scores
- Analytics overview – the full Reports suite, including team and call reports
- Disposition Codes – the Call Tags & FCR breakdown