How to apply AI at scale

© PATRICK FREY

Reading time: 20 minutes

Don’t try to change everything at once, but start with an important area.

Most CEOs readily acknowledge that artificial intelligence (AI) could revolutionize the way businesses operate. For example, they envision a future where retailers deliver bespoke products even before customers request them, perhaps the same day they are made – a scenario that may seem unrealistic, but that AI tools allow. already to materialize.

The problem is that companies have not yet figured out how to organize themselves to make this future a reality. Most of them have already made huge efforts to integrate digital technologies, in some cases going so far as to fundamentally change the way they meet their customers’ expectations and manufacture their products.

But to take full advantage of the promise of AI, companies need to reinvent their business models and ways of working. They cannot simply inject a dose of AI into an existing process to automate it or add useful information to it. It is certainly possible to deploy an AI system locally between several functions, for very specific uses (we then speak of use cases), but this is not enough to really boost the operations of a company or its turnover. ‘business. Such an ad-hoc approach also runs the risk of significantly complicating the larger-scale implementation of AI later on and increasing the resulting costs, because each team then has to start from scratch each time in terms of stakeholder buy-in, training, change management, data, technology, etc.

However, companies do not have to restructure all their activities at the same time to apply AI. This type of operation is almost always doomed to failure. Because the total overhaul of a company is an extremely complex process that involves too many variables, stakeholders and projects to quickly produce concrete results.

We have found that the right approach is to identify a critical business area and revamp it from the ground up. Modifying a basic process, journey, or function as a whole – what we call a domain – results in a huge improvement in performance that local one-off applications of AI simply couldn’t deliver. It also allows each AI project to build on previous ones, for example by reusing data or developing capabilities for the same set of stakeholders. We have observed that this approach gives rise to an organic cycle of change within the relevant domains and ultimately reinforces the desire to use AI throughout the organization, as leaders and their staff see it. efficiency. It also promotes the continuous search for improvement, which is vital because, due to the rapid pace of technological progress, organizations must consider AI-based transformations as a constant evolutionary process and not as a one-time project.

Eventually, companies that don’t take full advantage of AI will be overtaken by those that do, which we’re already seeing in several industries, such as automotive manufacturing and financial services. Fortunately, over the past few years, many companies (even those with limited data analytics capabilities) have begun to equip themselves with the skills to pull…

Magazine This article is for subscribers only, you still have 86% to read