Conversational AI: how it works

This first episode of Conversation Think looks at a hot topic of CX: conversational artificial intelligence and its applications. How do intent detection chatbots work? What are their limitations? How do they help improve the customer experience? We asked our very own iAdvize Machine Learning Engineers. Check out this video to get their answers.

The boom in e-commerce has been accompanied by an unprecedented increase in online queries. 

To respond to as many requests for information as possible, sites are turning to chatbots. The aim of these is to identify visitors’ needs and then direct them towards a solution.

A classic method is the deployment of a multiple choice bot, offering visitors as many options as possible... which leads to a far-from-natural conversation and an unsatisfying experience for the user. 

The perfect interaction is one in which the visitor can express their needs naturally. Conversational Artificial Intelligence lets you provide this natural dialogue.

Conversational AI is based on NLU technologies (Natural Language Understanding). The NLU engine analyzes the consumer's message to understand what they need.

It identifies:

- Intentions, which correspond to actions.

- Entities, which are pieces of information that specify the detail of these intentions. 

The NLU engine works in 2 phases.

- Training: it uses characteristic phrases to learn statistically which concepts are linked to an intention.

- Intention detection: the engine matches user inputs to learned phrases to extract likely intentions. 

It gives each one a confidence score, and selects the intention with the highest score. 

Many companies offer NLU engines. What are their strengths and weaknesses?

- They can spot grammatical differences between words in a sentence

- They can ignore common words. 

However, intention detection can be disrupted by:

- Spelling mistakes

- Very long sentences

- Multiple intentions in the same sentence

- The use of unknown synonyms, so that the list of intentions has to be expanded.

Overall, conversational AI has many benefits. It enables naturally flowing dialogues with users, and speeds up average resolution times. 

Conversational AI has many applications (IoT, messaging…). Gartner predicts that by 2022, 70% of customer interactions will involve machine learning applications, chatbots and messaging.

Sources: Caroline Collet & Jean-Baptiste Even, Machine Learning Engineers @iAdvize,, Dialogflow, Gartner, Top CX trends for CIOs to watch

Conversation Think

Speak your piece

Reach iAdvize’s engaged audience

Stand out from competition

Increase your brand awareness


Want to discover the solutions our platform offers?

Contact us