‍

We are entering a decade in which technology is becoming, among other things, a driver for the generation of new business solutions aimed at improving the customer experience with the brand through different communication channels.

According to Gartner, this year 30% of searches would be performed from screen-free devices, and 50% of all searches would be voice-based.

It is clear that the growing demand for these services, along with this type of forecast, forces us to study carefully their fit in our business.

Other sources indicate that 55% of teenagers use voice search on a daily basis, so if that’s our market target, we should take it into account in our strategy.

If we also take into account that the latest studies published by Juniper Research indicate that in 2019, 190 million euros were saved by the use of these technologies, and that cost savings are estimated for the next 3 years of 400 million euros in retail trade, 3.37 billion euros in the health sector, or 6.6 billion euros in the banking sector (equivalent to 862 million working hours)

In short, in view of these data, oral communication is very important for customers, and the cost savings for companies should be more than enough to keep an eye on these technologies. In fact, all companies, whatever the service they offer, which have a channel for assistance to the end customer, have a series of very elaborate procedures to ensure that their customers are treated as required and thus to consolidate the trust gained.

Voice processing has been using technology for many years, but the arrival of this technology to the end consumer has been possible with the arrival of voice assistants, mainly:

• Google Assistant and its Google Mini and Hub Nest devices
• Alexa and her Echo Show and Echo Dot devices
• Siri
• Cortana

When we want to use this technology in real use cases we have detected that the difference between the different types of assistants and which is the use case they intend to solve is not clear, causing the different Innovation teams to bet on this technology only as a proof of concept (PoC) that does not reach the productive environment mainly due to the following problems:

• The solution is not aimed at the target audience/sector (Customer Person)
• The technological resources for the implementation of a conversational assistant are not available
• Ability to analyze the impact of the solution on the target teams or customers

The different types of attendees that we have analyzed and that the community has accepted are:

ChatBots

They are assistants that allow human interaction with an application through a messaging system such as WhatsApp, Telegram, Slack.

The characteristics of this kind of assistant are:

• They don’t maintain a conversational context. They always follow an established linear flow
• They do not require the Impersonation of the user with other third party services
• Assistants based on “Action -> Response” interactions, e.g. “What’s the weather like this week?”

The benefits to the user are:

• Immediately
• Remove an interface full of buttons and menus. The need to learn how to navigate the interface disappears
• Ubiquity, because with a simple messaging app that includes the possibility of including bots we can open all the conversations we want with shops and services
• Accessibility, with a much more comfortable and intuitive interface
• Efficiency, the user gets more with less

ChatOps

The assistants called ChatOps are an evolution of the ChatBots, but oriented to facilitate the day-to-day operations of the work teams. It is a collaboration option that allows connecting people, tools and processes in a completely transparent workflow.

The characteristics of this kind of assistant are:

• They don’t maintain a conversational context. They always follow an established linear flow
• They require user Impersonation to use physical devices, because the device must have access to corporate credentials
• No Impersonation is required if you use a business application such as Microsoft Teams and Slack, because corporate credentials are used in the authorization and authentication process
• Assistants based on “Action -> Response” interactions integrated with the company’s interoperability solutions to facilitate the orchestration of an operation between different systems

The additional user benefits of a ChatBot:

• Access to corporate systems without the need for a computer. Most business solutions do not have an agile interface for mobile devices
• Increased productivity, for example to create the tasks to be assigned to our team at the end of a follow-up meeting

Conversational Voice Assistant

Conversational voice assistants are those that allow us to have a conversation between the person and the machine like if it was between persons.

These types of assistants are the most complete and the ones that require specialists to achieve a solution that provides added value to our customers/users.

The characteristics of this kind of assistant are:

• The maintain a conversational context
• They require user Impersonation to use physical devices, because the device must have access to corporate credentials.
• Conversation design based on conversation flows avoiding a linear conversation. That is, in other words, to go to step 3 we must go through step 2 and step 1. A conversation with a conversational flow allows us to reach a concrete action from multiple interactions.

The benefits to the client are:

• Immediately, it is not necessary to wait to be attended by an operator
• Optimise the physical network for user support
• To increase the “funnel” of attraction by being able to reach the greatest number of possible clients
• Provide an interface without buttons and menus. The need to learn disappears, as for example happens in an App or on the Web channel
• Streamline communication
• Authenticity, since we use what comes most naturally to us to interact in real life, which is language
• Accessibility, with a much more comfortable and intuitive interface
• Efficiency, the user gets more with less

To implement a voice assistant, the following points must be worked on:

• Definition of use cases and information required to define the conversational flows
• Governance of APIs, to expose the information of our systems enhancing productivity and user experience.
• Securing access to information systems through authentication and authorization processes
• Analysis to avoid identity theft
• Ability to analyze user interactions to improve response rate
• Access control and privacy of confidential customer information.

Additionally, voice assistants can be supported by Machine Learning solutions with NLP “Natural Language Processing” and NLU “Natural Language Understanding” engines.

NLP engines is a general term used to describe the ability of a machine to ingest what it is told, break it down, understand its meaning, determine the appropriate action and respond in a language that the user can understand.

While the NLU engines are a subset of NLP that deals with the much narrower, but equally important, side of how to best handle unstructured inputs and turn them into a structured form that a machine can understand and act upon. While humans are able to effortlessly handle bad words, words exchanged, contractions, colloquialisms and other peculiarities, machines are not as adept at understanding and assimilating those inputs with misspellings or other assumptions, for example dialects or Anglicisms that we use in our daily life.

But the voice does not just convey a message based on its words and context. The voice transmits a feeling that, if it’s correctly analyzed, allows our assistant to improve so that it can empathize with the client. For this purpose there are already “nascent state” solutions provided by Alexa Neural Text-to-Speech.

In future articles we will talk about how to face the challenges of implementing a virtual assistant in a company based on our experience and technical knowledge in the design of business architectures.