Help Center

Transfers and Retention

Best Practices: Conversation transfers

What is a transferred conversation?

AgentBot automatically resolves frequent user questions. During a conversation with a user, the bot is able to detect complex questions, or those that demand second level assistance (those requiring human intervention). These conversations can be sent to a representative or specialist—the question is transferred so it can be resolved by a human assistant.  This type of conversation, that starts with AgentBot and ends with a human, is considered referred or transferred.

When is it best to transfer to a human?

Questions are considered complicated or second level when they demand the intervention or attention of a person. This could be because of a particular need of the business or service, or because of internal procedures the company takes in the face of certain customer actions.

For example: returning products or canceling services, closing sales, responding to specific complaints, or a technical service that the bot cannot provide a solution to with generic instructions, etc.

What are the different types of transfers?

AgentBot offers two types of transferred conversations when it is necessary to refer:

1. Transfer because of answer. The question is transferred, because of company rules or business flow, and it must be resolved by a human. When the user makes this type of request, the answer will indicate that they will continue with a person, and will be transferred to the corresponding customer service area (technical, sales, customer service, or billing).

Some examples: 

“I want to cancel my plan.”  Most companies have a specific procedure to follow when a client wants to cancel or unsubscribe from the service. It follows then, that faced with these types of questions, the conversation will be transferred, so that it can be continued by a representative or specialist.

“I want to buy a new telephone.”  Some companies decide to close sales or upsell through a representative that is qualified to offer this type of service.

“There is no signal/it doesn´t work.”  The bot generally offers a series of steps to solve this type of problem, so that it should be resolved independently. However, in the case where the client has tried everything and the problem still continues, the company transfers the conversation to a technical service representative, so that they can look into the particular case and resolve the problem.

2. Transfer because of rule. This methodology is applied to subdue the user when confronted with negative behavior that can affect customer satisfaction.  In the text below, we explain the different guidelines that can be applied and how to avoid these situations.

What are the possible transfer guidelines? How are they applied?

Within transfer because of rule, there are three alternatives to rein in the user. You can choose all the rules, only one, or a combination of the three. The application will depend on the needs of your business, and also how developed the virtual assistant is in terms of the content it possesses.

1. Consecutive evasive answers. When the user receives consecutive evasive answers, the conversation is transferred. It is interpreted that the virtual assistant doesn’t understand the user or that the virtual assistant does not have the requested information. The number of evasions for this to occur is chosen by the client.

We recommend transferring the conversation upon the third evasive answer. This gives the user the chance to reformulate their question or resolve it based on the suggestions that accompany the evasive answer. Moreover, repeated occurrences happen when the user enters meaningless or incomplete text, which would not be considered a valid question to be transferred. 

2. Negative feedback. When the user gives consecutive negative feedback to every answer during the same conversation, the conversation is transferred. This is interpreted as the client not agreeing with the answers they have received. These could be correct answers, but they do not make the client happy, or they could be incorrect answers, if the assistant cannot interpret the question effectively. The quantity of consecutive negative feedback received for this to occur is chosen by the client.

We recommend transferring the conversation upon the second or third negative feedback. This leaves room for subjectivity in terms of the level of client satisfaction. It is common to receive negative feedback when the correct information is given—in this case, a human representative could probably not provide a solution other than the one provided by the virtual assistant. 

3. Same consecutive answers. When the user receives the same answer consecutively, the conversation is transferred. It is interpreted that the user has asked a different question about the same subject, but the assistant does not have any other information, so it gives the exact same response. Consecutive answers also occur with the user writes the same text more than once, thus prompting AgentBot to give the same answer. 

In this case, we recommend transferring the conversation before the third repetition.  This is also obviously the choice of the client.

The Methodology Behind Transfers

According to the strategy, influx of questions, and operating hours of each company, there are two ways to refer a conversation.

1. Instantaneous transfers. The conversation is transferred to receive an answer and is immediately attended by a human. According to the reason for the referral, it is transferred to the corresponding customer service agent, differentiated by specialization.

2. Delayed transfers. The answer offered to the user is to complete a form, so that their questions will be attended by a specialist that will contact the customer afterwards to help them. These forms will generally be integrated into a ticket system or addressed via email by the corresponding customer service specialist. According to the type of question, these types of referrals can contain various objectives for the company. For example: lead generation, database creation, a preliminary filter according to the client or subject, etc.

How do I reduce transfers by rule and increase retention?

Retention is the capability of the virtual assistant to solve questions automatically without referring conversations to a human. Increasing retention, by avoiding transfers by rule, can limit the number of users. This is key—their elevated numbers require the time of human resources that have been allocated to complex or second-level customer attention. To increase retention, it is important that the virtual assistant’s content is comprehensive, updated and constantly evolving. 


  1. Content is key.  Continuously generate content that covers all the needs and aspects of the business. This will limit the number of evasive answers and prevent the bot from repeating the same answers. Apart from having content, you should employ a logical method for generating questions, questions, ways to ask questions, and meanings involved.

    • When editing questions and ways to ask questions, or when you consider adding words within meanings, always remember to use the meanings contained in the AgentBot dictionary. This knowledge base evolves on a daily basis. A key point is to include the root word, so the most complete entry in the dictionary appears.

    • Take advantage of the Improvement section. Here you discover new issues that the virtual assistant hasn´t considered yet, or to learn words that your clients use that are not found in the dictionary.

    • Include more than one answer to evasive questions.  If you have various answers to evasive questions, the assistant can randomly alternate between them, which prevents user frustration and shows empathy. We suggest always including an apology and information that indicates that the Bot is learning, or information about the topics that it knows and can respond to at the moment, for example.

  1. Analysis of negative feedback. Keeping in mind that while user feedback is quite subjective, it can sometimes indicate that an answer is obsolete, incomplete or incorrect. Many times, however, feedback can in interpreted as the client simply having bad day or not agreeing with some aspect of the service.

    • From the Management section, the results are analyzed, ranking the content by the amount interactions and negative feedback received in the last 30 days. This allows you to improve the content that is most important to your users.

    • In the Tag Report, you can see the most requested tags and the percentage of negative feedback. This lets you improve the commonly asked questions in the last 30 days.

  1. Proactivity, integrations and actions. Integration with CRM and legacy systems is the key to providing useful information and solving problems in real time. Incorporating complements in the answers and integrating third-party applications will help to resolve users’ needs and improves user experience. These can resolve the question in the first contact and prevent the user from resorting to another channel, which generates multiple costs for the company for the same consultation.

    • Use the largest amount of complements possible in the answers: videos, images, FAQs, links, navigation assistance, etc. This will help to grab users´ attention, so that they better understand the answer. Remember that it is chat—the content should be adapted to this format.

    • If you take advantage of internal systems and integrate services, you can proactively meet the demands of your customers. For example: activate a product instantly, accept payments in real time, take orders or receive complaints, etc.

  1. Strategy and objectives. Retention is like as a water faucet, where the company has full control over the number of conversations that it lets slip by or retains.  A defined strategy to control the number of transferred conversations will help to increase retention.  The Project Leader can explain the best practices to assist you in defining and executing a strategy.

It is also important to have a goal or a desired value of retained conversations to guide the number and magnitude of the actions. For example: as an initial goal, aim to retain 40% of the total number of conversions and try to reach 60% in two months.

What to keep in mind when referring conversations?

Will the virtual assistant be integrated with live chat? If so, during the implementation process, keep the following points in mind:

  • Live chat hours of operation: Keep in mind that the virtual assistant operates 24/7, so if live chat has different hours, you must provide backup actions for the hours that live chat is not available. This could be, for example, a contact form or a message with additional information.

  • Messages for each situation: The user should always be informed of continuing actions. It is important that the messages the assistant shows before referring or when it is outside of the hours when a message can be referred, etc., coincide with the tone of voice of the company and with the content in general.

  • Customer Service Groups: If your company has different groups specialized by type of attention, let the AgentBot technical team know, so they can create the needed rules and tags. Inform your Project Leader which content should be referred to in each case, so that they can apply the corresponding tags.

  • Requirements: Give your Project Leader information about which human chat provider you use. They can analyze the technical requirements needed to configure and program the integration.

This website stores cookies on your computer. These cookies are used to collect information about how you interact with our website and allow us to remember you. We use this information in order to improve and customize your browsing experience and for analytics and metrics about our visitors both on this website and other media. To find out more about the cookies we use, see our Privacy Policy.

If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference not to be tracked.