Before choosing an Artificial Intelligence Assistant, companies must take many key factors into consideration.

Chatbot levels : How to choose your AI Virtual Agent ?

How smart are chatbots now? Is your bot as smart as you would like it to be? What makes your Virtual Agent smarter or less capable than your competitor’s? There has never been such a gap among chatbot solutions on the market and more challenges for enterprise decision makers to really know what virtual agent to pick.

At Konverso, we’ve been working on Enterprise AI Agents  since 2017, and we tackled numerous challenges ranging from Machine Learning and Natural Language Understanding, to Context Management, Personalization, Automation and Bot to Bot communication. 

Based on the analysis of successful implementations of our technology in the enterprise market, here is how we see AI agents  evolve over the next decade. We identified 5 levels of maturity, from basic FAQ chatbots to multichannel and multi-round conversational bots, pro-active, non-linear and even cooperative chatbots. 

See what type of artificial intelligence chatbot makes the most sense for your organization’s digital transformation now… and for the future.

chatbot levels : how to choose your AI virtual agent ?

Level 1: FAQ chatbot or single turn conversation

Most of us have become accustomed to basic bot capabilities, what we call “intelligence level 1” or the FAQ Chatbot. This is the most common type of virtual agent  currently used on the market, from ecommerce websites to public administrations online help. 

This assistant allows the user to ask a simple query to get an answer, which is a slight improvement from doing search in basic FAQ pages.

However, even in this category the capabilities of the different vendors already greatly differ. To clearly spot the difference among those vendors and platforms you can ask the following questions: can my virtual agent retrieve the answers in various knowledge bases, inside and outside the enterprise? 

How does my chatbot clarify the question when there is ambiguity in the language used? Is my chatbot pre-trained to manage semantic similarity? How does my chatbot manage the semantic beyond the basic features that a Natural Language Understanding (NLU) and Natural Language Processing (NLP) engine provides? 

How does my bot bring different responses based on the user’s role in the organization (an occasional user needs a different response extracted from a different source than an expert)?

Level 2: multi-round conversation chatbot

Some chatbots allow for multi-round dialogue, which means you can have a multi-step conversation with a basic level of back and forth. Furthermore, some questions can’t be answered in a single turn. 

It requires the chatbot to ask clarifying questions, without predefined scripting. In this category, the corporate bot can classify multiple intents (what the user wants to achieve) in order to select the next actions. 

Those conversations between a bot and a user often involve asking (prompting) the user for information, parsing the user’s response (a “human to chatbot” translation), and then acting on that information to retrieve the best answer (in the right language).

However, in this category also major gaps exist among vendors. Different trainings can be required for a problem solving chatbot to understand the many different ways users express their intentions. Because AI systems will have to deal with misspellings, synonyms, named entities, etc. 

Some chatbots leverage pre-trained deep learning algorithms and can be tested on users from day one, others need to be fully trained to be operational. You also need to check if your chatbot can recommend content based on user context and usage

If you already anticipate growing usages for your virtual agent , you might want to check if it can also leverage active learning to accelerate its autonomy and value for your organization.

Level 3: Contextualized / proactive chatbot

Highly conversational interface and context-aware chatbots can leverage user preference and other types of data such as usage and previous conversations in order to bring better services and to simplify the user experience. 

Those self-services are able to proactively engage the users and send notification to trigger a workflow. A chatbot connected to Enterprise applications will be able to use the data to understand issues affecting the end user. Based on data from the end user’s Analytics tool, the chatbot will pop-up to alert the end user that an action needs to be performed (run an automation, check for updates…).

Contextualization might not be on top of your shopping list now. But if you consider the importance of efficient interactions for user adoption and speeding up incident resolution, you start to see how it can positively impact the bottom line.

Level 4: non-linear conversation

People find it frustrating to use chatbots, because basic Virtual Agents rely on manually developed decision trees to handle “scripted” conversations. But natural conversations often go beyond the predefined and linear intent resolution paths created by developers. 

To further complicate this, users often fail to follow a task to its logical conclusion before initiating another one, which means that the chatbot must be able to guide the conversation in a new direction to achieve the desired result – without losing the original context of the conversation.

Effective context management is important because it allows bots to interact with users in a way that is easier, quicker, more helpful. This level 4 chatbot should demonstrate the ability to handle mixed initiative and tasks and take actions based on inference. Those robust bots can handle interruptions and switch context. 

Because users will not always follow your defined conversation flow, step by step, they may try to ask a question in the middle of the process, or simply want to cancel it instead of completing it.

Level 5: chatbot cooperation

In the near future, AI agents  will be able to know more about end user context and will act in multiple parts of a company’s operations—from IT Service Desk to marketing, sales, HR, or finance. We totally support this vision of how chatbot communication can evolve, but it requires collaboration between the players on the market. 

That’s why Konverso is an active member of the Alliance for Open Chatbot which defines interoperability standards to enable efficient communication among chatbots.

What is at stakes is a better user experience and in the long run creation of business value and innovation thanks to chatbot cooperation.  

What level of chatbot are you ready to build or buy?

 By 2022, 70% of white-collar workers will interact with conversational platforms on a daily basis (source Gartner). This dynamic for a chatbot-human collaboration is already in motion, but many vendors are still unable to deliver enterprise-grade chatbot solutions. 

You can decide to build your own chatbot if you have the right team and skills to deliver a high-quality level 3 chatbot. An alternative or even complementary approach is to look to third-party providers such as Konverso that specialize in pre-built AI Enterprise Assistant. 

Because collaboration works for chatbots and can also work between in-house teams and Konverso’s team of Machine learning and NLU/NLP specialists. That’s how we built a robust and evolutive chatbot platform, with a strong roadmap for the future.