Suggested Answers API

What are Suggested Answers?

Suggested answers can be shown to your customer service agents when they handle chat sessions with your customers. This allows them to respond to common questions and handle recurring steps in the dialogue with a single click. This saves time and money.

Using suggested answers can reduce the work effort for your customer service team with 10-20% right out of the box. This is something you can get for free.

You can also train the AI with your frequently asked questions. Higher automation levels can be obtained right away.


How it works

  1. Customer sends a message
    Your customer sends a message to your company via some interface (website chat, Facebook messenger etc.).
  2. Message is shown to agent
    The customer’s message is shown to your customer service agent.
  3. Message is sent to AI
    A copy of the customer’s message is also sent to an AI-module. The AI-module identifies the language and forwards the message to a range of different AI sub-modules.
  4. Suggested answers delivered to agent
    If one or more modules have a suggested answer that meet the quality metrics, these will be sent to the agent. The messages will appear on screen next to the message from the customer.
  5. Agent can send, edit or ignore
    The agent decides if they want to send one of the suggestions to the customer right away, or edit before sending or simply ignore and type something manually.



Different types of suggested answers

The suggested answers are generated by our different AI-sub-modules.

Each sub-module uses different data sources and different machine learning technologies to arrive at a suggested answer.

  • Generic answers
    The generic answers handle common parts of customer conversations, such as initial greetings, goodbyes and a few others. This works out of the box.
  • FAQ
    If you have uploaded an FAQ (Frequently asked questions) to train the AI, this can boost your automation level instantly.
  • Scraped data
    If you sign up for Sonetel with your business email, we will automatically scan your website and extract certain types of data. These will be added into a “data graph” for your business. If we detect questions related to any of that data, we will use the data graph to create a suggested answer.
  • Chat history
    The more you use the service, the more chat history data will be available to train our AI. If similar questions are detected later, this will trigger suggested answers based on historic chat data.