
Crescendo gives us a scalable first line of support that combines an AI assistant with human agents. The AI assistant is available around the clock and consistently exceeds our automated resolution target: it currently resolves over 65% of conversations on its own and peaked at 74% in one month. That has reduced the volume reaching our internal tier 2 squads and supports our goal of reducing workload with the same headcount.
The reporting is trustworthy. Conversation-level exports matched the figures presented in our business reviews, so we use that data for real operational decisions, such as sizing agent coverage by hour of the day. API access also lets us bring it into our own internal dashboards.
The team acts as a partner. When we raised issues with their support team, they gave us the attention we needed through to resolution. Their operations lead proactively brings proposals to adapt the setup to our actual volume patterns, Customer Success was flexible in structuring the contract around our roadmap, and as early adopters of their new platform we have a direct channel to their product team for feedback. Review collected by and hosted on G2.com.
The human first-response layer adds less value for a technical B2B product like ours. Its main contribution was a fast first reply, while complex questions still depended on our tier 2 team, and routing accuracy required ongoing attention from our side. This is part of why we are moving toward an AI-first setup on their new platform.
The AI’s automatic satisfaction scoring can be too optimistic. When a customer stops replying, the AI sometimes rates the conversation highly, as if the issue had been resolved, and in some of those cases we believe the customer simply gave up on the answer. Treating unanswered conversations as a separate category, or calibrating the score against direct customer feedback, would make this metric more reliable for quality management. Review collected by and hosted on G2.com.