Artificial intelligence, beyond experimentation

N°290a – Synopsis (8p.) – Technology
Artificial intelligence, beyond experimentation
Add to cartSubscribe

Artificial intelligence applications are multiplying in all sectors of activity. But many companies are struggling to develop beyond the experimentation stage. How can you reach a new level and take advantage of AI’s full potential?

Artificial intelligence applications are multiplying. Airbus uses intelligent drones for the maintenance of its airplanes: these devices, equipped with visual recognition tools, scan the planes and produce an anomaly report, shortening ground downtime. The French car manufacturer Groupe PSA has deployed a chatbot, Eva, which can provide immediate answers to administrative-type questions or to those concerning the everyday life of its staff. For its recruitment in China, L’Oréal is testing AI-assisted solutions to help identify appropriate candidates. American cosmetic brand Kiehl’s relies on AI to identify the best time to suggest, via email or SMS, to its customers to place a new order to avoid running out of their beauty products.

Yet, quite often, companies find it difficult to reach beyond the stage of an approach which aims at validating an idea, the PoC (Proof of Concept). They carry out multiple experiments but, in fact, AI only plays a marginal role in their performance. The reasons are many and varied: scattered initiatives, difficulty to generalize pilot projects, difficulties to integrate with standard operations, interfacing problems with existing information systems, reluctance of the staff to adopt the solution, etc. Thus, despite promising progressive steps, the return on investment often remains disappointing or, at best, uncertain.

Some returns of experience nonetheless help clarifying the approach needed to succeed in deploying AI. It emerges that the challenge goes well beyond technology, as underlined by the study conducted by Malakoff Médéric and Boston Consulting Group, Intelligence artificielle et capital humain. One of the authors of this study thus sums up the issue in an article in the French business magazine L’Usine Nouvelle: “People consider that AI consists in coding, writing algorithms, and that’s it! No, this is only 10% of the work. 20% must focus on making it work with the existing tools, and the remaining 70% concerns deployment, the work to see the processes adopted, to change the ways of working.” Here are the key success criteria that have emerged from recent returns of experience concerning large scale deployment attempts. 

SubscriberSign in
to download
the synopse (8 p.)

Sign in

Forgot your password?

VisitorI want to buy
this synopsis (8 p.)


VisitorI want
to subscribe