As the European Commission tries to regulate the use of artificial intelligence, there is no doubt that this technology is proving to be crucial for the economy of tomorrow. The most competitive companies will be those which have succeeded in putting AI and data at the heart of their business processes … Few have succeeded so far, the clouds of smoke having too often taken precedence over a profound transformation and methodical business.
Today, the majority of general managements and business departments have identified the use cases of data valuation that can serve their strategic and operational challenges. Most of them have already experienced some of these use cases.
The problem now is the scaling up and the effective integration of this data and these AI models into the core processes, the business actions and the underlying information systems.
Go from a “laboratory” logic to a “factory” logic
More and more large companies are undertaking the construction of data-AI factories to reconcile business needs, data science teams, BI teams and IT teams … in the service of the end-to-end industrialization of business projects for which AI and data become key components.
Quite simply, the idea here is to move from a “laboratory” logic (the famous data labs which have served well to incubate data science skills and trigger the desire and commitment of the professions) to a logic from a “factory” with efficient, end-to-end processes and tools to turn ideas and data into tangible business value. A path simple in appearance, but not without pitfalls:
– technological pitfalls to build professional data-IA platforms (security, quality of service commitments, resilience, robustness, versioning, real time, algorithm management, data management, collaboration, etc.) as well as “pipelines” (data pipelines ) necessary on the one hand to collect data from all sources (software packages, IoT, mobile applications, etc.) and on the other hand to display model results to user tools. On this subject, let’s start with the turnkey suites of cloud players (Amazon, Microsoft, Google, OVH) or on hybrid solutions combining in-house components and market software, skills are scarce and volatile and few companies have succeeded in building lasting robust teams on the subject;
– organizational and governance pitfalls because the subject is at the border between several departments: business departments who are potential users and responsible for data, data departments (data office or others), which have generally invested in data science and data governance teams, and IT departments, generally sidelined because of their “liabilities” (too slow, too expensive, not sufficiently agile and innovative …), even though they are often the most able to assume this new industrial logic (which they assume often already on the very close decision-making subject);
– funding pitfalls, these platforms and teams can represent high costs in Capex and Opex which are difficult to make profitable as long as the plant does not deliver enough and posing the eternal problem of the “pioneer payer” (the logic of financing bases being very often attached to business projects, the first projects are likely to finance the following ones).
Covid-19, de l’“IA glitter” at l’“Transforming AI”
Despite these many pitfalls, and undoubtedly helped by the health context and the digital acceleration it has induced, the phenomenon is growing and data-AI factories are multiplying in formats specific to DNA and governance. of each company (centralized, delocalized or hybrid, single or multi-publisher, global or vertical, outsourced or internalized skills, etc.).
The health crisis has revealed, sometimes painfully, the inadequacy of data strategies and governance, and especially the inability to quickly scale up ideas for improving or reinventing processes. This is particularly the case in the supply chain, deeply shaken by the crisis (shortage of raw materials, click and collect and new, more direct distribution models, new issues of sovereignty, more local production and traceability, etc.), this triggering numerous data-IA factory projects for the supply chain function. This is also, of course, the case for marketing and sales functions, which have seen an explosion in the desire or obligation to use digital channels by customers (13% growth in e-commerce in one year, according to a Kantar study), but have not always been able to respond with a sufficient level of service and personalization.
A recent Twilio study estimates that the crisis will have on average saved companies seven years in their digital transformation pace, but with very strong disparities between those who had or have actually invested in solid technological foundations (including AI factory type ) and those who have dealt with the most urgent without investing significantly. Finally, the profound changes induced by the crisis in working methods and the employee experience greatly accelerate the need for HR functions to better control and enhance the value of data in HR processes (loyalty, new management methods, absenteeism, mobility management and careers, training, etc.), leading to the emergence of vertical data-IA factories for human resources.
This movement is a strong signal: we are finally coming out of the era of “glitter AI” to enter that of “transforming AI”!
By Ghislain de Pierrefeu, partner at Wavestone law firm
Expert opinions are published under the full responsibility of their authors and in no way commit the editorial staff of L’Usine Nouvelle.
Source: UsineNouvelle – Actualités A la une by www.usinenouvelle.com.
*The article has been translated based on the content of UsineNouvelle – Actualités A la une by www.usinenouvelle.com. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!
*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.
*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!