G-1Q07VLPB62
top of page

Here's a run-through on how to develop AI Assistants by Rasa CEO Alex Weidauer.




AI (Artificial Intelligence) assistants have been in the market ever since Facebook Messenger

opened up their API, which allowed for the creation of chatbots. These chatbots interact with

users via simple texts and build up a conversation. However, when you start developing a

chatbot, you realize how difficult it is.


There are many AIs and APIs that support chatbots’ creation, but fine-tuning them to your

purpose is difficult and time-consuming. With the iPhone, a basic chatbot was introduced that sends a notification, which is a one-way communication with the user. But, over the years, chatbots have advanced to the extent where they can perform entire conversations with users.


AI Assistants and Rasa


An AI assistant or virtual assistant is a program that can understand human language and

commands, and perform the corresponding functions. These assistants can also respond via

texts or audio. The Google Assistant, Amazon Alexa, and Siri are examples of popular AI

assistants.


Rasa is an open-source framework for developing AI assistants. They invest in extensive

research that helps to create conversational AI. Developers make use of their cutting edge

Natural Language Processing (NLP) to understand the messages, determine the intent, and

capture the required information.


Dr. Alan Nichol and Alex Weidauer co-founded Rasa. They quickly realized that most AI

assistants and chatbots don’t perform well. Even those that work perfectly are built by a small group of people with machine learning expertise and access to significant resources. This is where Rasa comes in.


Rasa aims to help everyone make use of AI technology and build their own AI assistants so that they never have to use off-the-shelf AI tools by big companies.

The AI assistants built on Rasa can hold meaningful conversations with users, with multi-step

conversations that can remember the context and even integrate business logic. These

assistants are quite flexible and can be deployed on any platform, be it private cloud, on-

premises, or even a third-party cloud provider. They also can be connected to various existing systems, APIs, and knowledge bases.


Rasa is quite popular and trusted by leading companies worldwide. This is mainly because of

their increased security, HIPAA compliance and data privacy.


According to Rasa, chatbots can handle only basic questions and FAQs. Contextual Assistants are at the next level, and this allows for end-to-end automated conversations.


Let’s have a look at various levels of AI assistants.


5 Levels of AI Assistants


There are five levels of AI Assistants, and these are inspired by the autonomous vehicle.

1. Notifications

2. Chatbots

3. Contextual Assistants

4. Personalized Assistants

5. Autonomous Organization


1. Notifications

These are the most basic type of assistants that provide one-way communication with

the user via notifications. This was first observed in the Apple iPhone and quickly spread

to other devices and platforms. It could show up on a messaging application on your

device.


2. Chatbots

Chatbots are the most commonly available assistants in the market. By simple Natural

Language Understanding, these assistants can respond to customers and also facilitate

FAQs.


3. Contextual Assistants

Contextual Assistants are not easy to develop. With just NLU, you cannot develop these

assistants as you need to predict what happens next in the conversation. You do need to

understand what the user is conveying, however, you also need to focus on dialogue

management.


The context here refers to the context of the conversation and understanding this can

be key to a conversation for the assistant.


Rasa focuses on contextual assistants and makes use of deep learning and merges it

with business logic.


For implementing a contextual assistant, you require a whole team.


4. Personalized Assistants

A personalized assistant knows the user’s preferences, habits and choices and provides

suggestions or recommendations based on these.


5. Autonomous Organization

An autonomous organization is where an entire company functions on various AI

assistants. As a company, your customer service department might already be using an

AI assistant, this can be extended to include marketing, sales, and more. This is quite

complicated to implement and might be possible in the future.


Conclusion

Level 3 or Contextual Assistants are already being implemented with Rasa and other open-

source tools. But the higher levels are even more difficult to implement and may require

further technological development.


The 5 levels of AI assistants are useful for businesses as well as customers. It makes business

processes easier while improving customer engagement.


You can check out Rasa and their amazing open-source platform here.

Comments


bottom of page