We reviewed the transcripts and noticed that customers rarely phrase concerns as clear questions—they state issues directly and often provide backstory before getting to the point. The challenge with Smith.ai is that it seems to require customers to phrase their concerns as full, precise questions rather than recognizing key terms like "refund" or "password reset." If the bot can’t interpret natural language effectively, it may not be a scalable solution for us. Instead of reducing support volume, it could generate unnecessary transcripts and slow us down. We need assurance that the bot can filter out calls that can be self-served and only escalate when necessary. Are there planned enhancements for better keyword recognition, an option to receive transcripts only when follow-up is needed, or a way to bypass redundant questions like “Are you an existing customer?” With major campaigns launching soon, we need to determine if Smith.ai will streamline support or if another solution would be more effective.
