The adoption of Artificial Intelligence in the enterprise is advancing at great speed. But to be deployed on a large scale within organizations, AI must be “responsible”, i.e. it must inspire trust, be perceived as fair, unbiased, explainable and increase human capabilities without replacing them. Today, everyone is talking about responsible and ethical AI, there are more than sixty charters on the subject. But it is time to move from principles to action!
According to a study, the first ever on the level of maturity of companies in terms of responsible AI, less than half (48%) of companies that have successfully deployed AI throughout the organization have done so with responsible AI. Other figures are also telling: 55% of organisations overestimate their maturity in terms of responsible AI, but also almost half of companies that think they have successfully deployed responsible AI are, in fact, lagging behind. This study – “Do you overestimate your AI maturity?” – was conducted by BCG GAMMA, the AI and data science arm of BCG, among more than 1,000 large global companies.
Responsible AI is based on seven key principles: accountability, transparency, fairness and equity, data governance and privacy, social and environmental impact, safety and security, and the combination of AI and human intelligence. The companies surveyed are divided into four levels of maturity according to their progress on these seven dimensions: 14% of them are late, the majority are in the development stage (34%) or advanced (31%), and finally, less than a quarter of the companies (21%) are leaders on the subject.
However, there is a real gap between the perception of leaders on the maturity of their organization and the reality of their progress on the issue of responsible AI, and this for almost every stage of development. Data governance and privacy is the most advanced dimension for all. However, the dimensions of justice and equity, social and environmental impact, and the Human+AI approach (augmenting human capabilities without substituting for them) are the most difficult to implement and the ones where there is the least investment from organizations today. These are the dimensions for which Trusted AI could also enter the equation, as discussed in the third issue of our quarterly.
While there are major differences in deployment, these differences are also noticeable from one region to another. We note that Europe and North America are the most advanced regions with an index of 66.8 and 66.3 respectively against 62 for Asia or 60.7 for the Middle East. The difference between the perception of managers and the reality of deployment is very clear: 36% of European companies say they have fully integrated responsible AI but 52% of them actually overestimate their maturity on the subject. This is even more true for companies that have scaled up AI: 81% say they have fully integrated responsible AI but only 49% are actually mature on the topic.
It remains important to raise the existing opportunities for companies to move towards responsible AI. The main motivation for organizations to deploy responsible AI remains the business benefits (for 42% of them). This is even more true for the most advanced companies in the field. Then comes customer satisfaction (20%), risk management (16%), compliance obligations (14%). Only 6% want to deploy AI for social responsibility issues (6%). This is what Sylvain Duranton, Global Director of BCG GAMMA and co-author of the study, says in conclusion: “Today, the most advanced organizations in terms of responsible AI are no longer doing it only to better anticipate and manage potential risks. They see it as a real competitive advantage, whether it’s to differentiate their brand or to accelerate recruitment and employee retention. Implementing responsible AI is proof of a responsible innovation culture, one that is often supported by the company’s goals and values.”
Translated from Une étude de BCG-Gamma révèle les niveaux de maturité des organisations et entreprises quant à l’IA responsable
