Not all chatbots master the art of natural conversation. It’s one thing to make someone choose an option on a menu. It’s a different thing to maintain coherent dialogue with the bot, without limitations, just like you would talk with a friend on WhatsApp or Facebook Messenger.

Since there are so many technological options for customer service on the market, we’re letting you know the pros and cons of each one so you can make the best decision for your business.

Structured flows (without any intelligence)

Structured flows

 

Chatbots based on structured flows basically consist of an options tree or a structured if chain. In other words, if a certain condition is met, it executes a block of sequences.

It would be something similar to an IVR (interactive voice response). The user doesn’t have the option to write an open question but instead has to navigate to reach the answer.

What are the pros and cons?

  • Useful for answering up to 10 topics.
  • Ideal for companies that have less than 1000 conversations per month.
  • Easy to set up (onboarding).
  • Inexpensive.
  • Limited options.
  • Usually it can’t be integrated with other technologies or incorporate a human agent.

Natural Language Understanding (NLU) and Conversational Builders

Natural Language Understanding (NLU) and Conversational Builders

 

Within this group, IBM Watson and Microsoft Luis are the most popular. These technologies were born on the basis of understanding language, mainly commands, such as “I want to turn on the light.”

Generally, they require a structuring of intents (questions) and entities (synonyms or data), and a flow of expected cases.

What are the key aspects of these technologies?

  • Many of them have excellent features to intelligently take structured information and fill out a form or an order. For example, they’re useful for searching for a flight or ordering a hamburger.
  • Ideal for simple requests.
  • Some versions are On Premise.
  • Economical at the transaction level but expensive at the maintenance and component creation level.
  • They require predicting the way the client will interact in a flow.
  • They are viewed as one component. If you want to implement other features, such as interconnection to channels, statistics, satisfaction, complements, integrations, learning, scalability, etc., you must develop them separately.
  • The update can be slow and usually depends on IT staff.
  • If given a lot of content, it could fail. Therefore, it’s important to have a good team that manages the flow, intents and entities.
  • They usually don’t have spell checker and a database of regionalisms.
  • They may require another solution for this task or your team may need to manually upload errors, jargon and different ways of talking about the same idea.
  • They don’t keep context or memory. In some cases, this feature can be programmed.
  • These are multipurpose. Because of this, statistics don’t usually include specific details about feedback, satisfaction, channels, time, etc.
  • They don’t adapt the response to the channel being used, so you need different versions of the solution for each channel.

Conversational Artificial Intelligence and End-to-End Solutions

Conversational Artificial Intelligence and End-to-End Solutions

 

These options offer a comprehensive solution for a specific use. They include all the key components to meet the needs of the user and the company. For example, they offer connection to channels, human chat, learning tools, detailed analytics, data usage, among others.

These solutions also incorporate a conversational engine with multiple AI technologies for open and natural conversations. They’re usually built on linguistics, semantics and meaning.

They easily understand verbal conjugations, gender and context thanks also to technologies such as NLU, Machine Learning, clustering, recommendation engines and others.

They’re usually oriented to customer service or eCommerce.

Their main features are:

  • They correct mistakes, distortions in language and meaning.
  • They understand context and short-term memory.
  • They understand emojis and voice messages.
  • It’s not necessary to structure the conversation. Conversation is open and based on questions and answers.
  • They offer a human chat solution or integrations with other chat applications.
  • Statistics are detailed and specialized by use.
  • They have a learning tool.
  • They can be managed by non-technical staff. This is specifically useful given the fact that content changes so frequently.
  • They have greater precision with more content.
  • Some have persistence features. This allows the conversation to be kept on different channels.
  • Some have security and data privacy certifications such as GDPR. This aspect is important when handling sensitive information such as with banks, for example.
  • They offer integrations with other customer service solutions.
  • They adapt the answers to the channel automatically allowing an omnichannel experience.
  • They have a complements and self-management base, such as carousels, buttons, ticket generation, flows, etc.

Our recommendation?

Before choosing a technology, define your strategy. Identify how you want to interact with your customer or user, what your volume is, and what you’re able to do to maintain your bot. After that, you can evaluate the tools that will help you achieve your goals.


Related article: Technology at the service of customers or customers at the service of technology?


Selecting the right technology, understanding how your customer wants to talk with the company and long-term thinking are the keys to creating an incredible experience, improving satisfaction, saving costs and increasing sales.