AI agents have revolutionized customer service, with projected consumer spending on purchases made via bot interactions expected to reach $142 billion worldwide in 2024, compared to just $2.8 billion five years ago.
As AI agents increasingly drive business performance, the need to ensure quality in bot interactions becomes paramount.
Zendesk QA now offers comprehensive quality reviews for AI agents, enabling you to measure, track and improve your bot service quality effortlessly.
Visually separate and manually QA bot conversations
New AI agent QA functionalities help you identify and focus on conversations handled by bots, saving significant time and effort required to sift through the vast amount of tickets and data.
Conversations handled by Zendesk AI agents (native and Ultimate) are automatically identified as AI agent interactions. Use the Participant, Bot, Bot type or Bot reply count filters in conversation and assignment views to find conversations handled by AI agents.
To gain a deep understanding of your bot performance, review your AI agent conversations manually. Select the relevant bot as the reviewee, choose the right scorecard for assessment, and rate conversations as you would for regular agents.
To streamline your QA workflows, set up your Zendesk AI agents under “Users, bots, and workspaces”, separating them from the regular user management.
Gain full coverage and critical insights into AI agent conversations
Understanding the overall quality of your bot interactions and identifying issues are crucial for refining and training your AI agents.
Zendesk QA’s Spotlight automatically detects critical issues in AI agent interactions, presented in the conversation view. You can also filter AI agent conversations based on the following Spotlight filters:
- Bot repetition: Detect conversations where the AI agent repeated the same message or question multiple times in a row.
- Bot communication efficiency: Find conversations where the bot handled conversations more efficiently than your average agents – faster and with fewer questions. Or, instead, less efficiently by exchanging more messages than human agents when handling similar customer issues.
- Sentiment detection: Identify escalated conversations in which an AI agent’s response caused positive or negative customer sentiment.
In addition, AutoQA automatically scores all AI agent conversations in the Greeting, Readability, Spelling & Grammar, Solution, Tone, Empathy, Closing, and Comprehension categories, providing you a thorough overview of your bot service quality.
Note that AI agent conversations are currently not measured using the AQS metric, and bot quality statistics are not included in the AutoQA dashboard. You can find the manual QA results on the Reviews dashboard, and dive into the new AI agent QA (BotQA) dashboard.
Learn more about AI agent QA here >
Explore the Bot QA dashboard
AI agent performance and quality data are now neatly presented in the new Bot QA dashboard. Filter data by time period and specific bots; focus on bot-only conversations, or escalated tickets.
- Track the ratio of your Bot-only conversations vs Escalated conversations and how it changes over time.
- Monitor the Escalation rate over time to notice changes in the number of bot conversations escalated to human agents.
- Check how often the Bot repetition, Bot communication efficiency and Bot sentiment detection Spotlights are identified in your AI agent interactions, and how they change over time.
Use Zendesk QA for an effortless and consistent quality program for your AI agent interactions. Keep a close eye on your bot performance and ensure a high service quality at all times.