When evaluating the agent experience, keep in mind the FAQs.

To train your agents, start by enabling recommendations for a single skill.

In Conversational Cloud, inline recommendations (answers and bots) are displayed directly within the conversation.

Two recommendations being offered to the agent inline in the conversation

  • Click Delegate to join a bot to the conversation, so the bot takes over. You stay in the conversation, so you can monitor the bot’s progress and remove the bot if needed.

    The conversation flow when a bot is joined to the conversation

    Tip: As shown in the image above, a system message announces when the bot joins the conversation. You can customize this message.

  • Click Use Answer to copy the article’s text to the agent’s text input area. You can edit the text before sending it to the consumer.

    Use answer button for using a recommended answer The conversation flow after a recommended answer has been used

Provide feedback on recommendations

You can provide "thumbs up" and "thumbs down" feedback on recommendations. Give a "thumbs up" when the recommendation is right based on the consumer's query. And give a "thumbs down" when it's wrong.

Here's how to give feedback on an inline recommendation:

An agent giving a thumbs-up for a recommended answer and then sending that answer to the consumer

And here's how to give feedback on a recommendation that's offered in the On-Demand Recommendations widget:

An agent giving a thumbs-up for a recommended answer and then sending that answer to the consumer

Learn how to turn on this feeback feature.

Remove or replace the current bot

After you have joined a bot to a conversation, you can remove or replace it if desired:

  • To remove the current bot, click Remove bot at the top of the messaging panel. The agent can then take over.
  • To replace the current bot, click Replace bot beside the bot you want to substitute into the conversation. The selected bot joins the conversation, taking over for the previous bot. (Only one bot can be joined to a conversation at a time.)

    Remove bot and Replace bot options that are available when a bot is a part of the conversation

Notify the agent when the bot has finished

If you’re recommending bots to your agents, it can be a challenge for the agent to know when the bot has finished its work. The agent must check back repeatedly on the bot’s progress. To solve this, the bot can send a private message when it’s finished handling the consumer’s request. The private message can tell the agent what action has been taken, and let them know that it’s time for them to rejoin the conversation to close things out with the consumer.

Look up answers and bots on demand

Conversation Assist automatically recommends answers and bots to agents, inline in conversations, based on consumer intent and conversation skill. But…sometimes…your agents need more flexibility. Sometimes, they need to be able to look up answers and bots on demand, regardless of what the consumer just said. The On-Demand Recommendations widget in the Agent Workspace meets this need.

An agent using the Bots and Answers tab of the On-Demand Recommendations widget to find bots and answers

If you’ve turned on the display of the widget, you can use the Bots & Answers tab to ask any question, or enter a phrase, and get back available bots and answers. You can then easily use those recommendations in the current conversation.

The Bots and Answers tab of the On-Demand Recommendations widget

Note the following identified in the image:

  1. Copy answer: Copy the recommended answer (the plain text) to your clipboard in order to paste it somewhere else.
  2. Edit and send answer: Copy the recommended answer to the conversation window, where you can edit it before sending it.
  3. Send answer: Send the recommended answer immediately.
  4. Delegate to bot: Delegate the conversation to the recommended bot.

As with recommendations that are displayed inline in the conversation, all recommendations shown in the widget depend on the skills assigned within Conversation Assist. For example:

  1. The agent picks up a conversation on the “Ordering” skill.
  2. The agent uses the widget to search for available answers and bots.
  3. The results returned include only answers from knowledge bases assigned the “Ordering” skill and only bots similarly assigned the “Ordering” skill within Conversation Assist.

Other info in the widget is also skill-based. For example, if the agent’s active conversation is on the Ordering skill, the widget’s “Most used by all” list includes only the answers and bots used the most in conversations on the Ordering skill.

Delegate a bot on demand

In the On-Demand Recommendations widget, the Most used by all and Most used by me areas offer, as expected, the answers and bots that are most used.

When delegating a conversation to one of these listed bots, the agent must specify the message to use to trigger the correct business flow in the bot.

Enter the message to use when delegating the conversation to a bot via Most used by all or Most used by me in the On-Demand Recommendations widget

This ability to specify the message to use makes a big difference. Consider the following conversation:

  • Consumer: I heard about your Internet/TV offering from a friend. Can you tell me about it?
  • Agent: Sure, we offer the fastest speeds and all the popular TV channels. Cancel anytime. Interested in signing up?
  • Consumer: Yes, please
  • Agent: {Agent looks up bot in Most used by me in widget}

Ordinarily, the consumer’s most recent message is used to trigger a business flow (i.e., a dialog) in the bot. But as the conversation above illustrates, this won’t always yield the right initial response from the bot when it joins the conversation. This is because the most recent message from the consumer (like the one above, which is “Yes, please”) might not be intentful.

When delegating a conversation to a bot on demand, being able to specify the message to use to trigger the correct flow in the bot solves this issue. In our example above, the agent could enter “sign up” when delegating the conversation to the bot. And this would trigger the “sign up” business flow (dialog) in the bot.

Train your users on the keywords and phrases that trigger the bots used in your solution.

Edit a rich answer

You can edit the text (only) of rich answers that are offered via Conversation Assist.

Here's how to edit a rich answer that's offered inline in a conversation:

An agent editing the text of a rich answer that's offered inline in the conversation

And here's how to edit a rich answer that's offered via the On-Demand Recommendations widget:

An agent using the Bots and Answers tab of the On-Demand Recommendations widget to find a rich answer and to edit the text of that answer

Look up replies on demand

Predefined content is a set of canned responses (replies) for common use cases: greetings, closings, and so on. Conversational Cloud lets you personalize predefined content, so it reflects your brand’s voice and business needs.

Predefined content is made available on the Replies tab in the On-Demand Recommendations widget.

An agent using the Replies tab of the On-Demand Recommendations widget to find predefined content

If you’ve turned on the display of the widget, you can use the tab to search and browse for replies on demand. You can then easily use them in the current conversation.

The Replies tab of the On-Demand Recommendations widget

Note the following identified in the image:

  1. Copy reply: Copy the reply to your clipboard in order to paste it somewhere else
  2. Edit and send reply: Copy the reply to the conversation window, where you can edit it before sending it.