Chatbots, real conductors of your data

In the space of a year, chatbots have been invited to all debates. Now, if they prove relevant for certain use cases, reducing the subject of bots to chatbots, would be to look at the world by the small end of the spyglass. We are going to deal here with bots in the broad sense: the progress of the AI comes to multiply the tasks that one can entrust to them.

A bot is a software brick with a specific mission. We are far enough from Wall-E! What defines this little robot is not his appearance, but his role. Here is a simple example: your answering machine is a bot, because it responds to your external solicitation. In fact, as soon as a software takes care of a job for you, autonomously, we can talk about bot. In other words, we are surrounded by hundreds of bots; and the phenomenon is not new. What is, however, is to want to use them to transform the business and take the digital turn.

Strong AI and weak IA

One of the founding fathers of Artificial Intelligence, Marvin Lee Minsky (1927-2016) defined his discipline in the following way: "Science whose goal is to make a machine perform tasks that man accomplishes using his intelligence ". AI is usually divided into two parts: weak AI and strong IA. The latter, long reserved for science fiction, is an artificial intelligence capable of doing the same things as the human being, while being aware of herself. But we are starting to hear some experts saying that we will have strong AI, maybe 30 years from now. Or ... a thousand years! The projections are still extremely variable.

In the meantime, the weaker AI offers a field of possibilities largely overlooked by companies, with the exception of GAFIM, which have taken a big step ahead. The weak AI already performs feats, with of course the victory of Deep Blue (IBM) against Garry Kasparov, during a game of chess in 1997, but also, since, victories in poker or the game of go robots that learn to bluff and become unpredictable. Autonomous cars, voice assistants, AI able to play the piano, paint canvases or film, but also explosion of the IoT, 100% connected store, without checkout ... examples swarm.

However, the most interesting is perhaps elsewhere, far from these emblematic examples: in the way that AI can transform the daily tasks of companies.

Alignment of the planets

To take other examples that we all know, the ability of couriers to detect spam offers us a comfortable service. When our iPhone rings and an unknown number is displayed, we appreciate that the smartphone suggests to us who called, reconciling the caller with a number that will be found in an e-mail signature. These functions are not perceived as intrusive, but on the contrary as relevant: the bots are then precious auxiliaries.

Transposed across a company, this type of service opens several avenues: AI can be used to automate many tasks.

So, when a marketing director is thinking of launching chatbots, it would be best to put all this in context: what businesses really need is mostly a bot, not a chatbot . Chatbots benefit this year from an "alignment of planets" driven by the explosion of the social media, but we do not necessarily need a conversation tool with the client, rather than a tool for optimizing operational.

Humans love "gray areas"

Deciding to entrust a certain company's tasks to a bot requires first of all a deep reflection on how the company is structured. For example, there must be a clear grid of products, which is far from obvious, from one agency to another or from one region to another.

Humans love "gray areas". But the machine does not understand what is not formulated. As soon as one decides to automate, one raises unprecedented questions on the processes. - Christophe Tricot, Big Data & Data Science Manager Kynapse byOpen.

Then, to rely on a bot asks to formalize a number of processes that have never been. What diagnosis does this employee make before advising this or that insurance product, to this particular client? Nobody asks these questions because it is not in human nature to explain everything. Humans love "gray areas". As soon as one decides to automate, it raises new questions, and it becomes even violent, because it forces individuals out of soft areas "soft" and natural.

We saw this recently with the deployment of bots at Credit Mutuel (Watson): it was initially poorly with the unions.

New skills to acquire urgently

It is believed to start with a customer relationship bot, a "little chatbot" to unclog customer service, but a small chatbot, it does not exist ... If the company does not lead a deep reflection on its processes, on what she brings to the customer, why and how, so she creates a chatbot above ground that is only a "wart", a showcase, in fact a totally counterproductive tool.

To be able to lead this thinking, companies need specific skills. Because, as has been said, self-analysis is a violent operation. There is no suspicion of the extent or scope of what is tacit in a company. But machines do not understand what is not formulated. Companies rarely have the necessary skills internally. And there is no offer on the shelf. It's still the "Far West" ... We can ask a machine to analyze three million interactions with customers, but that does not mean that we can identify good practices.

More than mathematicians or data scientists, companies therefore need "facilitators" that is to say profiles able to go to the trades, to discuss their methods, their needs, their shortcomings ... They map the competences of the different services (with a constantly changing raw material) and establish an absolutely necessary dialogue. There is no job description for this new job, which revolves around the famous "gray zone".

Anthropologists and developers "plumbers"

To work on this gray area and on the acceptability of the bots (the reception they will receive internally, as well as from the general public), companies need to engage ... anthropologists. A feature still very rare in their offices, but which has all the skills required.

A second type of profile is needed, this time to immerse your hands in the construction of the bot: you need developers-integrators with a very particular state of mind. Indeed, they manipulate tools that potentially recognize the characters, the type of documents that are presented to them, the feelings expressed, etc. It's a huge toolbox and you have to have the ability to handle and integrate these different bricks. They are "plumbers" who juggle with semantic, cognitive technologies and also more classical technologies.

The skills required are rather the temperament of each. This is good news, as there is no academic training! These new behavioral skills can not only be found, but especially develop in business. We are entering the era of creativity ... It is up to companies and HR in particular to detect and nurture these profiles. The role of HR will be central and also requires a profound transformation. We look for intrapreneurs, "hackers", to form ideal pairs with the bots.

Internal mobility will be the key topic for years to come. Without adequate human resources, no one can hope to succeed in experimenting with AI.

An example: a company employs about twenty people to sort e-mails by hand every day: to forward them to the service concerned, to answer them, to delete them if necessary, etc. A bot is quite capable of understanding an intent in an email and performing this sorting. But nothing prevents reassigning the employees concerned to new missions, by giving meaning to the work of each. The skills people need to get the most out of AI are not only learned on school benches. Each company has the choice, between restructuring without thinking further, or inviting its troops to commit to its side, to develop the right know-how and appropriate bricks AI.

The use cases of bots have few limits: it is up to you to decide what to do with them. A virtual assistant, who takes your appointments and replies to your e-mails for you, as does Julie Desk, who has just raised 2.5 million euros? A bot that can search for you additional information on the web before you give a financial product to a customer? A bot that plunges into your editorial archives, extracts articles published ten years ago and contextualizes the information you needed? Everything is possible.

A shortened Time to Market

Companies are interested in artificial intelligence with very different approaches, some advance forcefully when others, especially pure players, start from start to finish and progress very quickly.

Each company has the choice to restructure itself without thinking further, or to invite its troops to commit to its side to develop the right know-how and get the bricks of AI. - Christophe Tricot, Big Data & Data Science Manager Kynapse byOpen.

New players can occupy the field at any time, with more effective solutions than yours. The current context of explosion of AI generates tension, the anxiety of not being at the rendezvous, as the fear of errors by excessive precipitation. In reality, you do not really have a choice: you have to go fast. The safest is to launch POCs quickly, to confirm a belief, to expose your ideas to the only valid justice of the peace (the client) and to choose the tracks that will be developed first.

Incoming data and outgoing data

And on the side of big data? Bots are also renewing the field of possibilities, with two main types of data: the ones they produce and the ones they inject.

The first category is rather new for companies: if we take the example of a chatbot that generates conversational data, we obtain a wealth of information very rich in lessons. Compared to the number of information that "comes down" from a telephone set, the bot delivers an exhaustive, intact raw material that can be subjected to a semantic analysis. This improves the service over time, but also to learn live information about the person with whom we interact: is it relaxed? Benevolent? Annoyed? Potentially lying? On the verge of breaking his contract? And so on. The vocabulary used, the syntax, the intonation ... everything can be sifted and indexed in a sort of thesaurus of emotions. Another possibility which is based on a strategy "old as marketing", the introduction of CRM data in bots. Analyze the user journey, its friction, its "hot spots" becomes easier.

With a bot, it is easy to keep everything and treat everything, but companies do not always seize this opportunity because they do not make it a priority. Still, if you had to give two tips, it would be the following: first, collect everything you can. Of course, this comes at a cost, but it is certain that tomorrow you will benefit from this untapped data. This is a gold mine that will allow you to build models of hyper-segmentation. Then watch. We saw the Microsoft bot become neo-Nazi in a few hours ... Keep your hand, improve your bot continuously, do not let it escape you.

This article appeared in  Digital Marketing 2018 was written by: Christophe Tricot of KYNAPSE OPEN.