What Is a Good CSAT for AI?
A good CSAT for AI support is one that holds even with human-handled tickets, generally in the 85 to 95 percent range for a strong team. The critical move is to segment CSAT to AI-handled tickets rather than reading a blended score, because a bot that resolves quickly but coldly can drag AI satisfaction far below the human baseline while the aggregate still looks fine.
The bar is parity with humans
Gartner data puts chatbot-only CSAT at 28 percent against 82 percent for human agents, which is the gap most AI support tools quietly live inside. A genuinely good AI CSAT is not "better than nothing," it is close to what your human team scores. If AI-handled CSAT is 20 points below human-handled CSAT, the automation is not ready for those tickets.
Tone is where AI CSAT dies
The most common cause of low AI CSAT is a bot using FAQ voice on an angry, urgent, or grieving ticket. Empathy is engineerable: surveys show most customers believe AI can be empathetic and want it to adjust tone to sentiment. The failure is not that AI cannot be warm, it is that ungoverned bots do not read the room.
How Resolve protects CSAT
Resolve collects CSAT on both AI-resolved and human-resolved tickets so you can compare them honestly. A sentiment guardrail keeps an angry ticket from ever auto-closing, and every answer is grounded in your knowledge base and matched to your configured tone. Low-scoring resolutions feed back as candidates for knowledge base improvement rather than being ignored.