Get access to all the tools and information necessary for the customer's life cycle at Odigo by heading to our client portal
To design a high-performance chatbot that is able to provide the best possible customer experience, the most important thing is to make sure it is well-supported. By mobilizing a wide array of skills and professions, you can ensure that your conversational agent is part of a global journey that nurtures its ability to understand natural language (NLU).
In the wake of the big data wave, it is time for the AI wave. The advent of chatbots has led to the creation of new skilled jobs. Before they can operate, conversational agents must first be designed and programmed. Once in operation, they need a human hand from time to time, no matter what people say. Here is an overview of the skills and professions necessary to create a high-performance chatbot.
For a chatbot to be effective, three good fairies must lean over its cradle: efficiency, personality, and eloquence. Their joint actions will be based on significant teamwork, involving a variety of actors from diverse professions.
To function properly, the chatbot must have a well-made “head.” The project manager (or project team) must, first and foremost, define its missions, its scope of intervention, its way of interacting with other devices, and the journeys the chatbot could be part of. This project manager will have participated in the technological choice of the solution beforehand – and it is likely that his choice will have been influenced by both ease of use and maintenance, particularly in terms of language understanding and restitution.
He is also responsible for supervising all KPIs measuring the success of the conversational agent, while keeping in mind the question of its integration with the rest of the customer environment. In this way, he ensures that the conversational agent:
To maintain the chatbot’s good performance, the project manager is usually assisted by an ergonomist. His role? Define personas, i.e. user profiles with a focus on how they expect to be served.
The way in which the service is provided, and the different cases possible are the subject of scenarios. They will have some things in common like be activated according to the user’s choices, and will apply decision rules according to his profile (his persona) and business rules (customer versus prospect for example). What is critical at this stage of design is to imagine all the possible branches. The ergonomist will configure the scenarios with the decision rules that the project manager determined.
The chatbot also ensures that the user experience (UX) is as pleasant as possible and enhances the interaction with “talking,” “visuals,” and/or entertaining moments. To do this, it can use carousels, checkboxes or geolocation of points of interest. The objective: a chatbot that is able to provide a relevant answer as quickly as possible!
Let’s not forget that the chatbot can often be the first point of contact with the brand for a virtually infinite number of Internet users. It acts as a showcase. The way it expresses itself, the look and feel of its avatar, but also, and above all, the content it is able to produce must reflect its brand identity.
It is the designer/copywriter who gives it its personality. The “bot persona” will indeed vary according to the sector of activity. It should be rather serious, even austere in the banking or financial world, conversational and empathetic in the worlds of insurance or retail, and could even share jokes with certain audiences!
To do this, it will have to adopt situational vocabulary, phrases (friendly or using a neutral pronoun), and specific styles (serious, sustained, light, sympathetic; using smileys). It must be pleasant and create a complicity with its audience to make the user want to engage with it, and come back to talk to it again.
Ensuring that the chatbot is able to speak (well) gives it all its relevance. This is where the linguist comes in. Often the same person, the designer/copywriter puts on his linguist cap when writing the chatbot’s pool of answers. The common misconception that a conversational agent can come up with its own answers is wrong. It is based on a library of questions and answers. The more varied this library is, the more the answers will be original and allow you to have a pleasant conversation, but the most important factor will be the chatbot’s ability to understand.
The linguist will have the task of teaching the chatbot to understand. To do this, he will train it with a corpus (a set of sentences from the users’ requests and their association with a pattern), then familiarizes it with the users’ multiple formulations. His mission will be to regularly feed its corpus to enrich its understanding. He will supervise its learning throughout its life.
When a conversational agent does not understand, it is because, in reality, he does not know. The request is new in the sense that he cannot associate an intent with the formulation. Quite simply, the chatbot was not trained on this verbatim because the request may be:
Let’s add that a user adopting an ironic tone will almost systematically confuse our conversational agent…unless it is supervised! In this kind of situation, our bot trainer acts as a “prompter.” Able to interpret the pattern instead of the chatbot, he immediately intervenes to continue the conversation without the user even noticing.
Since the bot trainer ensures that the conversational agent takes into account new requests (provided they are frequent and representative), its language is enriched and so are its missions! A chatbot whose only mission at the beginning was to track parcels can, after some time, take care of complaints as well.
For your bot to be effective, it must process volume, which may require switching to multiple channels. If it is given the gift of ubiquity, it must not lose its singularity! It is also important to be aware of the existence of technical limitations: what is available on one interface may not be available in exactly the same way on another (for example, on the website, the chatbot may allow a video to be viewed, while on Facebook Messenger a link will redirect to the video).
Duplicating the chatbot on a website or mobile application implies staying in the digital domain. Why don’t we take the plunge and move on to the voice media? It will be easier to start with a voice application (voicebot), where the uses are close to chatbot on messaging, but where the interaction is conducted by voice command! Some adjustments will have to be made, because we do not speak the same way we write. The ergonomist and linguist will have to adapt the language to get to the point faster. This will be an opportunity to prepare for a completely different use, this time on the phone channel (callbot).
To ensure the best possible customer experience, the important thing is to think of the bot in a global journey, without neglecting the means to maintain your natural language understanding (NLU). This requires good support. Natively integrating with the Odigo Contact Center as a Service solution, our chatbot will provide the personalized service you have imagined for your customers on the channels they prefer (website, mobile application, Facebook Messenger, etc.).
Interested? Request your demo here.
Reducing contact center wait times is foundational to improving customer experience (CX). Customers are abandoning brands that fail to respect their time, and switching to companies that find ways to continuously reduce contact center wait times. Deploying the right solutions positions contact centers to gain new customers and build lasting loyalty.
The new model of customer experience (CX) in banking favors digital channels over traditional in-person interactions. So are bank advisors a thing of the past? Quite the opposite. Thanks to new customer experience management tools, advisors in banking services are providing more support than ever, and embodying the trust that customers expect.
As a consequence of digital transformation, more and more contact center managers are turning to artificial intelligence (AI) and more specifically to one of its subfields: natural language processing (NLP). Read on to learn 3 ways NLP helps in increasing productivity and delivering memorable customer experience (CX).