Artificial intelligence: beyond the buzz, a major challenge

Artificial intelligence (AI) applications are in full swing. Long limited to research labs and science fiction, they have nowadays become a concrete reality for companies and their customers. Google uses Deep Learning, an AI technique, to perfect some of its key products such as Search, Gmail, YouTube or Maps. Virtual assistants such as Siri, Cortana and others are omnipresent in our computers and smartphones. Watson, an AI software developed by IBM, assists numerous companies in their customer relationship management. At Crédit Mutuel bank for example, it analyzes the 350,000 daily mails from customers, classifies them according to their urgency and, for some, prepares a standard response that account managers will then personalize. At Bouygues Telecom, Watson interacts with the clients through a chatbot—a software robot that can dialogue with individuals through automated conversations, in natural language?to help them connect their internet box.

Noting this impressive progress, numerous experts are suggesting that we are at the dawn of a “new industrial revolution”. Leaders are themselves convinced. According to several surveys, 85% consider that AI could bring them a competitive advantage?or help them keep the lead they already have. And the same proportion plan on investing in AI in the two to three coming years.

However, experts are warning against the “rush for AI” phenomenon. As for any new tool, we must identify its relevant uses, as well as the optimal operating conditions. For example, AI requires a huge volume of data. It also requires that human staff members cooperate with devices that are suspected, rightly or not, of taking their place?which is no easy task. Finally, it appears that AI’s added value is largely conditioned by non-technological factors. Three success criteria emerge in particular:

- Ensure you clearly define the boundaries of the fields propitious to AI in your company.

- Put in place an accompaniment framework to facilitate the acceptance of “intelligent” tools.

- Ensure an ongoing maintenance of the algorithms.