What you should know before launching your chatbot – part 1

First and foremost, you should know how to differentiate between two types of chatbots.

• Some are simple chatbots that work on a declarative system, which limits conversation-based interactions while providing the user with a simple and logical system. These chatbots work based on decision-making trees, which replicate the hierarchy of a conversation.
• Others are data-driven and predictive chatbots which rely on complex AI, allowing them to learn. These “intelligent” chatbots can understand (almost) any human request and formulate a response based on the intention and context of the request.

Step 1: identify business challenges that chatbots will help to overcome

For a bot to provide convincing and engaging experiences, you must first define its target(s) and the need(s) that it will meet.

A chatbot is not intended to – and will never be intended to – respond to all user needs, as this could result in convoluted messages that will confuse users. Users have everyday problems that should be prioritized. Do they need help finding a store, or booking a train ticket or hotel room? Do they need to be put in contact with customer service in a faster and more simple way?

Each feature that is developed and incorporated in future chatbots should be considered a way of responding to an existing need. A potential error here would be creating a chatbot with an unreasonable number of features too quickly and without enough study. A chatbot should be simple, and this simplicity comes from establishing a clear and specific goal before consciously limiting its capacities. Incidentally, it is important to remember that chatbots are vehicles of conversation; however, conversation requires language and thus chatbots’ limits must be set so that they do not crumble under the weight of translation and linguistic nuances therein. 

Step 2: define distribution channels

When we talk about “a chatbot”, we’re talking about the platform itself. This choice represents to a certain extent the image we’ve created beforehand. The choice of chatbot development platform will depend on the target (age, socio-economic background) and the need (business objective). Here, it is important to remember that a chatbot can be textual using messaging apps, for example, but it can also be vocal.

Below is a non-exhaustive list of possible platforms:

• Facebook Messenger
• Google Assistant
• Alexa (Amazon)
• Siri (Apple)
• Whatsapp
• Website or mobile app
• SMS

Facebook Messenger: brands’ preferred platform

This decision is not to be made lightly and, as with the functions, such a project cannot rely on an exhaustive list of platforms at the risk of getting lost. Replicating a chatbot across platforms might sound simple enough, but each platform differentiates from others by particularities of use and functions that change user behaviour with  the bot. For example, Facebook bots can include carrousels, and Google Assistant lets users interact with maps.

Step 3: choose a personality that fits with business and marketing objectives

Engaging with clients and providing a personalized and authentic experience is the key to success. Bot personality is one of the key aspects of chatbots, as it will be a reflection of your company – a kind of brand ambassador.  This personality or persona should be unique. While it is true that the user will not think of a chatbot as a “friend”, he or she will be speaking directly to the bot personally and privately, in an organic way. These conversations will help establish a privileged user/brand relationship (if your chatbot has a clearly identified personality!).

This personality should have specific characteristics including name, tone (familiar, casual, or formal) and even a visual component such as an avatar or profile photo. Finding the right name for a chatbot is a fairly difficult task. Should you choose a name that is typically female, male, or neutral? What’s more, most surnames are already being used. If you’re thinking of naming your bot Tom, Dick or Harry,, think again! All of these are already copyrighted. 

And don’t forget to get your marketing and communications teams involved in these discussions (if it’s not already the case!) as the bot will be part of the dialogue with consumers. 

Step 4: define the conversational flow

In a way, this is about deciding how the bot will respond to pre-defined needs, and what conversational tactics it will use to respond to user needs. Conversational flow should be considered as a mirror to specific user needs. This of course leads to conversation patterns that will be more or less closed and complex, which should be identified in order to be efficient in the development phase. 

Example of conversation flow

Step 5: put the user experience first 

You cannot choose your platform without considering UX. Generally speaking, the advantages of each platform should assist and guide the user in his or her conversation with the bot. The cases the bot handles result in conversational flows that can be guided and driven by the platform’s UX features. For example, Facebook Messenger allows users to select pre-defined responses which simplifies the discussion between bot and user.

These responses will drastically limit errors that could have occurred during the discussion if the bot had not understood the user’s request. This also reassures and encourages the user to continue the discussion, as he or she does not have to write an answer. Messenger also allows scrollable carrousels, illustrating each response option with photos.

It is important to take advantage of the various features that each platform offers, as it is often a good way to make conversation as fluid as possible.

A few things to keep in mind

You should remember that a chatbot is an ambitious project that must be firmly outlined and planned for. Recognizing a tool’s limits as well as its technical and functional capacities is vital for the success of the project. The chatbot must create harmony and not cannibalize existing resources. Always include all stakeholders in the project discussions, no matter the extent of its impact on their work. This will help existing teams to welcome the project and will contribute to a smoother launch (especially in terms of communication).

There you have it – all the fundamental questions you should be asking before kicking off your chatbot project. In our next article, we’ll tackle the more operational aspects of launching a chatbot. What are the implications for marketing and communications teams?  And, most importantly, how can you use data to measure performance? Sign up for our newsletter to be sure you don’t miss a thing!