Introduction

The superiority of AI virtual agents over chatbots is undeniable. The main difference between these two solutions is the technology used; the first use AI and statistical methods, while the latter are keyword-based.

Chatbots are rule-based, which means they “scan” messages for specific keywords in order to provide an answer. If they fail, they fallback and enter an infinite loop of rephrasing requests or transfer the user to a customer support representative.

AI virtual agents on the other hand, use machine learning and supervised learning to understand the meaning of a sentence. They can be quickly trained to comprehend what the user desires, using a few examples the business provides. Hence, they can accurately understand sentences with unknown words and typos, providing an elevated experience for the end users.

For example, the business trains the AI virtual agent with a few example-questions regarding the business hours. e.g. When are you open? What time can I come? If a user asks “What time should I drop by?”, a chatbot will fallback while an AI virtual agent will understand the meaning and provide the appropriate response.

Self-learning AI virtual agents

Just like humans would work, but faster.

However, not all AI virtual agents are equally smart and capable of improvement. Many use supervised learning, but very few use unsupervised learning, aka self-learning. And here lies the future of conversational AI.

Self-learning AI comprises of deep learning algorithms capable of self-improving an AI virtual agent. How? By analyzing large amounts of data and detecting patterns. AI virtual agents equipped with this capability can be a lot more than a customer support agent. One of the greatest business challenges is to understand what your market needs. A lack of that knowledge can create a gap and subsequently, bad customer service and experience.

With self-learning, you can bridge the gap between customer needs and business offerings.

Let’s say you deployed an AI virtual agent to handle a specific set of tasks.
How do you know your users are not asking questions that your AI virtual agent is not trained for? And if you know your AI virtual agent’s knowledge is limited, how do you select the appropriate tasks to “teach”?

Here is where self-learning is vital, in order to maintain a truly intelligent AI virtual agent that can support your operations!

Instead of spending countless hours manually uncovering new customer needs and integrating them to your AI virtual agent, allow self-learning algorithms to do the work for you. How? By gathering thousands of data from past conversations and understanding the FAQs that your AI virtual agent has no knowledge of. The end result? A continuously improving AI virtual agent with minimal human supervision that will always answer your customer’s ever-changing questions.

With self-learning, businesses are entering a new era of automation and are able to see the true potential of conversational AI.

How Moveo.AI works

We use machine learning algorithms that make use of Large Language Models (LLM). Our algorithms “browse” through thousands of chat logs, in order to identify new ways of improving the AI virtual agent’s understanding of how people think and communicate.

We use supervised learning, during the creation of the AI virtual agent. Our users create a few categories (Intents) and examples with customers’ most burning questions. In the case below, if a customer asks “how can I get in touch with you?”, the AI virtual agent will understand what the user desires and provide the business’s contact details. Pretty straightforward.

However, what makes us unique is our self-learning capabilities.

We use unsupervised learning in order to improve the AI virtual agent’s performance and keep up-to-date with customer’s ever-changing needs. Self-learning runs in the background gathering all the questions. Depending on their nature, it can either suggest new examples to existing Intents (new ways customers express themselves) or entirely new Intents (question categories).

For example, let’s say you are a Financial Institution and you know your customers usually contact you to ask where the closest ATM machine is located. As a result, you create an intent category named “nearest_atm”.

Auto AI will gather all customer questions and cluster them into groups. As a result:

1. You could end up with new examples of how customers communicate their need to find an ATM machine e.g. “close ATMs”. This will allow your AI virtual agent to better understand how your customers express themselves, avoiding fallbacks and clarifications. Better understanding, quicker service.

2. You could end up with entirely new needs such as “I wanna get a new visa card”, thus adding a new question category (Intent) named “get_new_credit_card”. This will improve your understanding of customer expectations and thus your services. Better understanding, improved customer satisfaction.

Indicative Use Case

How the Greek Consulates of UK and New York adapted to Covid-19 questions.

Our partnership with the Greek Consulates in the UK and NY began before the outbreak of Covid-19 pandemic and during that time, most of the user’s inquiries revolved around passports. However, when the pandemic began, a flood of messages started with questions regarding vaccination, certificates and entry to Greece requirements.
Self-learning took care of everything as within a few weeks, it clustered all questions and created new Intent categories regarding the pandemic, such as “accepted_vaccines” and “entry_requirements_greece”. All the Consulates had to do was accept the suggestions and provide an answer to the questions.

Our Human-In-The-Loop approach

We believe that the ultimate customer experience is achieved through a continuous feedback loop involving human and artificial intelligence. Throughout our platform, the “human touch” is imperative, from creating the AI virtual agent, to accepting the Auto AI recommendations and providing the answers to all the new Intent categories. AI is not a substitute for humans, but rather supplementary, in order to create outstanding experiences.