Organizations are accelerating their adoption of AI

Artificial intelligence has been gaining ground for several years now. Projects are multiplying: as proof, "in 2020 more than 80% of companies have experimented or adopted Machine Learning, a form of AI, compared to 55% in 2019." In addition, 63% of companies using AI in their supply chain management have seen an increase in revenue. AI also enables the automation of many routine tasks, and 69% of these are expected to be automated by 2024.

The 4 main AI approaches

From standard to customized, companies do not all have the same use of AI, sometimes constrained by the limited budget: 

• Use business intelligence tools: without representing an additional cost for the company, these tools allow you to analyze and advance predictive dimensions thanks to the data for an almost instantaneous update of the dashboards.
• SAAS solution coupled with AI: This is a solution developed for a specific context or sector: chatbot, optimization of building energy consumption, automation, or analysis of contractual documents... The uses are multiple.
• AI accelerators: concerning the accelerated implementation of use cases. Thanks to Machine Learning or APIs, AI allows us to develop use cases based on algorithms in order to test use cases and not to start from scratch. 
• Customization: Customized AI projects offer to precisely meet professional needs. Their nature is vast but they can include innovative processes or optimize existing techniques according to specific specifications.

AI: the major issues to be addressed

5 challenges :

1. Establishing its business strategy: Before including AI in the company's projects, it is necessary to evaluate its external environment (behaviors and trends related to its business sector) as well as internal (company's capacity: tools, skills ...). Finally, it will be possible to define what added value AI can bring.
2. Need for new skills: the integration of AI in a company requires new skills in this field but also an understanding of the stakes on the part of all the teams for an optimal use.
3. Governance, roles & responsibilities: each business function of a company must not only understand the actions related to AI, but also feel involved in its development. The ultimate goal is that everyone adopts it to make it evolve and apply it to their actions.
4. Cultural transformation called "change management": technological innovation is not only a question of application but also of acceptance. "You have to ask the professions where this AI could be applied and invite them to project themselves, this favors acceptance and adoption", says Eneric Lopez, Director of Artificial Intelligence at Microsoft France.
5. Ethics: in order to adopt AI on a large scale, it is important to ensure that the issues surrounding it are clear and respected, starting with transparency.

It is possible to make the data say many things. It is important to know where the data comes from because the more you scale up, the more biases will be felt in the results. 

- Luc Vo Van, Solutions Architect Data & IA at Microsoft France

Two approaches for scaling up

“Upscaling: for organization purposes, this requires alignment between its strategy, AI culture, business value and operational model."  Two approaches to do this:

• The programmatic approach: This holistic and operational model is often developed thanks to an innovation unit that centralizes all AI needs. This is a first step to acculturate collaborators, identify use cases and make everyone want to apply it on their own scale. However, this innovation cell must be decentralized later in order to have AI managed by the people whose job it is. At this point, the organization made aware by the innovation cell will be able to adopt it easily. 
• The modular approach: It consists in applying AI to specific use cases from the beginning. Each division holds its own AI needs and skills. "This implies having defined an AI strategy upfront applied to the different divisions or products of the organization." AI is then applied independently to each project, but the whole thing must be part of an overall AI strategy to ensure consistency across the organization.

In conclusion, in my opinion, the industrialization of an AI project requires the harmonious combination of three aspects: technical, organizational and cultural. If one is missing, industrialization will not be complete.

- Ygal Levy, Division Council Director of Microsoft France