Search

How to write scripts for your chatbot


What defines the quality of your bot? Cool features? Emojis? Variety of channels? No. Messages. Messages that your bot sends to your leads and customers. They can make or break the overall impression of your company. So how do you create messages that provide value both to recipients and your business? To do that, you need a system. But before I share the details about the system that we use, I'd like to go through the values that chatbots bring to businesses.

1. The conversation is an opportunity to exchange values with your customers and prospects directly.

2. The conversation is a natural and intuitive communication interface that everyone knows, and removing the human ego from there makes a positive impact on sales conversations. 3. Easy to scale human-like sales conversations. No need to hire, train and manage salespeople.

Based on these values, we developed bot messaging principles: CONTENT. The message is a solution for the user and helps achieve his goal in no waiting time.

CONTEXT. Provide personalized solutions on the user's specific request based on the data from previous interactions.

SUCCINCTNESS. No more than ENOUGH info to fulfill the user's request max 160 characters per message. Be clear. Avoid ambiguity and obscurity.

URGE. Fulfill business purposes. Move the conversation forward with questions till: a. User goals reached b. Bot task completed c. Conversation ended.

SOCIAL INTELLIGENCE. Be responsive even if you can't provide a solution. Every conversation should end gracefully, whether the user (and bot) tasks completed or not. So how do you do all that with the system? People interact with bots differently, and how can you prepare for all possible situations and scenarios? In reality, there are only five possible situations in in-person communications, and you can be prepared for all of them.

1. SIMPLE REQUEST. Bot understands what the user wants. So it can respond accordingly. 2. IMPOSSIBLE QUERY. Bot understands but can't provide a solution. For example, the user requests an appointment on Sunday at 3 am. In this case, you have to have a message that informs her about working days and hours and asks to choose another day or suggests an alternative. 3. MISUNDERSTANDING. Or, in other words, when the bot can't understand the request. In this case, you need to ask the user to reformulate the question. For example, if our COVID-19 bot misunderstands date and time, the following message is triggering, "sorry, which date and time were that?". Sometimes people ask completely irrelevant questions like which pizza can I order? Here you should remind the user about the purpose of the bot saying something like, "I can help you to check your symptoms and book an appointment with a doctor. Now how can I help you in this context? 4. NO RESPONSE. It occurs when the user left the room, is distracted, or maybe changed their mind. In this case, the point is to take advantage of this component and provide closure in a way that demonstrates social intelligence, but before that, give the user chance to re-engage. For example, in our COVID-19 bot, we use 3 prompts if there's no response to the question about a convenient date and time for the phone consultation with a doctor. The first one follows up in 3 minutes, "so which date and time would you prefer?" Two minutes later, the second prompt gives another chance "If you are still interested in a phone consultation, provide preferable date and time." And two more minutes later, bot closes the conversation by saying, "No problem, it seems that you don't have time right now. Feel free to contact us anytime. We are available 24/7. Bye." 5. TECHNICAL ERROR. And the last possible case that you might come across during the conversation is a technical error. On websites, sometimes you see 404 error. You have to be prepared for something similar. For example, a calendar app connected to the bot doesn't respond to a query. In this case, send something like, "I can't help you right now due to technical issues. Would you prefer to do it another day?" This approach guarantees that your bot never stuck and always gives relevant responses and closes the conversations appropriately. You can check how this system works for Calen.ai.

20 views0 comments