AI-powered live chat software combines two things: a chatbot that answers recurring questions automatically, and an interface where your team takes over when human assistance is needed. Whether this combination succeeds in daily operations depends on the handoff between the two. Eight criteria help guide the selection: context-rich handoff, routing, a unified inbox for all channels, returning conversations to the bot, CRM integration, data privacy, analytics, and the licensing model.

The Eight Criteria

1. Context-rich handoff. When the chatbot transfers a conversation to your team, the complete history must be included, along with the information already gathered by the bot. Otherwise, the agent has to ask everything all over again, and the customer perceives the chatbot as a detour.

2. Routing. A spare parts inquiry belongs in technical support, an invoice question goes to accounting, and an English inquiry goes to an English-speaking agent. Good software routes inquiries based on department, language, and workload.

3. A unified inbox for all channels. Website, WhatsApp, Instagram, Messenger: your team should view and manage all conversations in one central location without switching between tools.

4. Returning conversations to the bot. Once an issue is resolved, the conversation can be handed back to the chatbot—for instance, for confirmation or the next procedural step. Software that only hands off in one direction ties up your team longer than necessary.

5. CRM integration. During the conversation, your team should be able to see what the CRM knows about the customer and store the outcome there afterward. The guide to live chat with CRM integration describes which details belong in the customer profile.

6. Data privacy and hosting. Where are conversations stored, who can access them, and how long are they retained? Are inquiries sent to an external AI provider? The same questions apply to live chat as they do to chatbots.

7. Analytics. Which inquiries end up with the team, and why? Analyzing handoffs reveals where the chatbot lacks knowledge and where a specific process is missing.

8. Licensing model. Many live chat tools charge per agent. For teams with shift work or seasonal peaks, this quickly becomes expensive. A flat-rate price independent of team size keeps costs predictable.

Evaluating the Criteria in Practice

A demo shows what the software can do. Testing it with your own inquiries demonstrates how it performs in day-to-day operations. Three test scenarios are enough to gain a solid first impression:

1. An inquiry the bot cannot resolve. Check what reaches the team: the full conversation history, the captured information, and the reason for the handoff?

2. The same inquiry across two channels. Does it land in the same inbox, with the right team, and linked to the same customer profile?

3. Returning the conversation after resolution. Does the conversation return to the bot, and does the bot know what was discussed?

How Mercury.ai Combines Live Chat and Chatbots

In the Mercury.ai Agent Desk, your team takes over conversations from the chatbot complete with the full chat history and pre-qualified information—routed by department, language, or workload. All channels converge into a single inbox, response templates accelerate recurring replies, and notifications can also be sent directly to Slack if desired. Agents can take over, pause, and end bot dialogues. Following the live chat, the conversation returns to a predefined point within the chatbot.

The Agent Desk features a bidirectional integration with Salesforce. Teams already using Zendesk or Userlike can also transfer conversations there with full context. Analytics reveal which conversations were handed off and why. The Agent Desk is offered at a fixed price, regardless of team size, with unlimited seats. It is hosted in Frankfurt, and conversations are never used to train models. Learn more on the Agent Desk page.

Would you like to evaluate these criteria for your own customer service? Get in touch with us.