It was the end of the 2008 crisis that accelerated the process of digital transformation. In the wake of the GAFAs, traditional companies saw digital as a major competitive advantage, rather than just a technological evolution. As far as AI1 is concerned, mysteriously enough, a number of surveys show that the people questioned are fully aware of the stakes, but still seem to be waiting to see before believing.
Several books and articles describe the milestones that contribute to the success of this type of project. We have cross-checked this information with the experience of several managers and project leaders who have led transformations, to extract a list of key success factors.
1. Define the process to be improved and set quantified targets
First and foremost, you need to determine what you want to improve. AI is not the objective, but the means to achieve it. This starts with the definition of a high-stakes use case in which the processes are well mastered. AI brings performance, it doesn't solve organizational problems.
The return on investment (ROI) of this implementation needs to be closely monitored, both in terms of performance and speed. An application that reduces tedious, recurring tasks to enable employees to perform higher value-added tasks should deliver results in less than a year.2
2. Provide a clear, unifying vision
All the advice on digital transformation is clear: it's vital that the company's top management is involved in setting out a clear vision. Indeed, digital transformation generally implies organizational evolution and new ways of working.
In the same way, managers need to be fully aware of the issues at stake and the resources to be deployed in order to get their staff on board, and to ensure communication in order to :
- highlight the potential of digital in implementing the strategy,
- play down change by showing that AI is not intended to replace humans, but to assist them by reducing errors or handling repetitive tasks,
- prepare for the operational changes that will inevitably arise (new skills, training, team reorganization, new business processes, etc.).
3. Data: fuel for artificial intelligence
In most projects, the first step is to transform the company so that it is oriented towards the needs of its customers. data is crucial. A significant proportion of AI implementation projects that struggle to make progress are because they have encountered problems with data quality or labeling, leading to a lack of confidence in the models. To feed its AI algorithms, an organization needs data that is available, complete and relevant to their business.
At the same time, by processing a large amount of data in this way, whether personal, commercial, technical or industrial, the organization must address the risks linked to cybersecurity, confidentiality and evenethics, which also introduces a security requirement.
4. The dream team
The team leading the transformation will play a crucial role. As with any project, it will be necessary to integrate multidisciplinary resources, each with their own area of expertiseexpertise. Often in-house to understand the business and represent users, and external for the functions of data-science and development functions responsible for creating the solution.
Increasingly, successful organizations are training one or more employees familiar with the company's core business processes in the principles and uses of AI and machine learning algorithms. They will then be able to project AI into their business. 2
The team will benefit from being grouped together on the production site, for a better grasp of business issues and better communication, and led by managers applying best project management practices.
5. Preparing your organization
It's an illusion to think that some kind of "plug and play" miracle solution will come along and ensure that results are optimal. All the managers we interviewed know that the road to parameterization and implementation is long and difficult. A long testing and ramp-up phase is often necessary, and the result can be abandonment.
In fact, the many changes brought about by this transformation will all take place simultaneously, bringing with them an additional risk.
Three quarters of successful projects were prepared with the teams. Whatever the size of the project, the project managers involved all the users from the outset. They were pleased with this later on, as a lengthy fine-tuning phase was required, during which the level of pressure and activity rose.
6. Scaling up, large-scale deployment
More than any other project, AI needs to be implemented progressively. After collecting business data and generating the rules needed to program and train the AIs, experimenting with the algorithms on a test case can demonstrate the return on investment. The third stage is the deployment of solutions across the entire organization. There are 4 main reasons why it is so difficult to scale up AI:
- large-scale data recovery and formatting,
- adapting the algorithm to work with the new perimeters,
- the infrastructure, which will have to adapt to the new volumes of data being processed,
- confidence in the model, which is often harder to obtain from people who have not experienced the transformation from the outset.
Once the project has been deployed, however, the transformation is not over. Managers and their teams are often faced with new challenges: new organizations, new skills, AI reshuffles the deck.
>> Read our next article on " Artificial Intelligence: three challenges for managers and their teams ".
By Philippe Merckling
Read two inspiring books on digital transformation to prepare for your project:
1.Emily Métais-Wirsch & David Autissier, La transformation digitale des entreprises : Les bonnes pratiques - Axa, Pernod Ricard, Sanofi France, Schneider Electric, les Echos Ed. 1. (Eyrolles, 2016).
2.Stéphane Roder, Guide pratique de l'intelligence artificielle dans l'entreprise : Anticipper les transformations, mettre en place des solutions Ed. 1. (Eyrolles, 2019).