1. Think “User Experience”

When our customers embark on a chatbot project for the first time, they imagine a robot that will be able to manage all questions, learn by itself and manage all requests automatically, even those they have not identified themselves in their FAQ or their support processes.

Obviously, this is not the case: a chatbot is a graphical interface or a mobile application, that must have a dedicated user experience. It is a conversational experience, very different from browsing a large FAQ database or querying keywords in a search engine. However, it’s still a human interaction with a machine. But this machine, regardless of the promise, won’t be able to do more than what it was programmed for. So it is imperative to remember all the best practices of UX design: target personas, typical paths, guiding the user with buttons or carousels…

Despite considerable progress in machine understanding of natural language, the user must be accompanied through their experience. They use the chatbot for a specific need and must, above-all, find an answer to their problem as easily as possible.

  1. Targeted Use

Consequence of the previous point: a chatbot must be dedicated to a specific use, and this must be made clear to the user. A train ticket reservation chatbot must not also manage the support (claim, refund, information, etc.) for example. In addition to the targeted use, you must also clearly identify the intended audience and how best to address him.

If you want to do too much with only one bot, you will confuse your users and risk significantly reducing the quality of project by not having in mind the priority use of your bot.

  1. Think content creation

A chatbot is a communication agent of your brand and company. Its way of expressing itself, the content of its answers, the images and the links it will broadcast are elements of communication. They must correspond to the use, as we have seen, but also to an identity of its own. Do not forget to define the bot’s identity and its level of language: how it speaks and it behaviour. Is it polite, direct or relaxed in the tone of its messages…?

Integrating communication skills into your chatbot project is a point that cannot be undervalued.

  1. Plan Improvements

The ideal chatbot that will learn on its own and improve effortlessly is a myth: no successful chatbot works like this. Artificial Intelligence in the case of chatbots is there to better understand the intentions of the users. To know how to answer correctly, it will pass through either a database of knowledge in which it picks the answers to submit to the user (with a precise Semantic Analysis of the contents), or in content which you have precisely defined according to different scenarios.

In order to improve it, it will be necessary to plan its “maintenance” will be organized: as for a human, if we do not explained what to do and where it is wrong, it will only reproduce the same mistakes…

It’s common sense but disappointing: we cannot believe that AI will learn completely on its own, any more than an advisor can learn a new process, a new feature of your website, or a new regulation without we describing said function to it.

So always plan the maintenance load!

  1. Allow back up plans

A chatbot aims to better respond in real time to the most frequent requests: it is effective on the 20% of topics that are requested by 80% of users. But for the remaining 20% of users with more complicated needs, it is sometimes necessary to be able to speak to a real person. Allow users to get in touch with competent human advisors who can handle the unexpected. Whether by chat, email or phone, a well-designed chatbot must offer support by an advisor.

Discover & download our last ebook about chatbots! (article and ebook actually in French, soon in English)


Also discover the interview of Léonard Pommereau, R&D Project Manager at Cap Digital, about the ChatBot Studio.