Analytics: Clarity on 
every conversation

Delight.ai Analytics is built into every AI agent conversation and covers resolution rate, handoff rate, CSAT, flagged message and hallucination reporting, data exports, and more, all updated in real time.

Resolution rate measures how many conversations your AI agent closes on its own. Handoff rate measures how often it passes a conversation to a human instead. A rising handoff rate on one topic is usually the fastest signal something needs attention.

Every conversation transcript in delight.ai Analytics shows sentiment, tags, resolution classification, and every knowledge base or tool call the AI agent made, alongside what the customer and AI agent actually said.

Flagged messages are customer messages caught for a monitored topic, like a request to delete an account, or an attempt to override the AI agent's instructions, so a person can review them before they become a bigger problem. You can also define your own safeguard categories beyond the four built in by default.

Yes. Low-confidence AI responses, the ones most likely to be hallucinated, get flagged for review in the same place as flagged messages, so a bad answer gets caught instead of repeated. Dynamic Safeguards can also generate a context-aware response in real time when one triggers.

Yes. The Data Export API returns conversation-level detail or daily rollups as JSON or CSV, including message logs, sentiment, and resolution outcomes, with PII masking built in, straight into your own BI or reporting stack. Read more in the Data Export API announcement.

Delight Agent MCP lets you ask about resolution rate, handoff rate, or a flagged conversation directly from the AI tools your team already uses, drill from a metric into the conversation that explains it, and update agent instructions, safeguards, or handoff rules without opening the dashboard.