What is Innovation@scale? Take a behind-the-scenes look at the group's latest meeting on AI for forecasting.
In this 30-minute replay, find out how Michelin successfully deployed AI in its Supply Chain for 15 stores in Europe and 4,000 different products. Focus on machine learning for long-term, item-level production and supply forecasting.
Dive into the heart of an Innovation@SCALE meeting
The Digital and Technologies Lab has set up a working group called Innovation@SCALE. This team is made up of members in charge of driving innovation in the Supply Chain in their companies. Every 6 weeks, for 1 hour, they talk about their successes and failures in scaling up.
To better understand what it means to share experiences in industrializing scale-up, we decided to film an extract (half) of the last meeting and give you access to it via a replay with timeline. Today, the site is a forum for exchanging use cases for inspiration and improvement.
Since its creation in February 2023, Renault, Deret, Geopost and Michelin have presented their innovation governance, i.e. how they steer the large-scale deployment of their projects, but also the more forward-looking part of innovation.
Understanding why and how Michelin deploys AI for its Supply Chain.
The presentation begins with a definition of the department's activities by Pierre Cordina, Supply Chain/IA leader at Michelin and today's speaker. This department works on Artificial Intelligence, Business Intelligence (BI) and advanced analysis reports, even proposing solutions to humans.
Michelin deploys Artificial Intelligence (AI) at various stages of its Supply Chain so that machines can make autonomous decisions for people. In this video, we focus on machine learning to forecast tire production and supply requirements.
Forecast error reduced by 5-10%, equivalent to 1-2 days' inventory
As of 2019, this so-called tacticalforecasting solution provides long-term monthly forecasts for factory staff. It helps determine the quantity of tires to be produced (for countries with a factory) or imported (for countries without).
The D@ril platform, a key factor in the success of the project's international deployment
Encouraged by the success of an initial machine (s@@m) that detects stock outs in the network, and suggests and corrects situations, Michelin's teams applied the same 6-step approach to scale up the forecast solution:
- Collecting data
- Data cleansing
- Teaching the past to the machine
- Shaping the machine
- Testing the machine
- Predicting the future
Phase 5 is a double runi.e. the machine is tested, while human predictions continue in parallel. This is done in order to compare conclusions, but above all to facilitate adoption by operators. In fact, on the basis of tests on a growing number of items and good results observed by the teams, confidence in the solution is built up.
Today, a machine is always coupled with a forecast manager, i.e. an operative. They are very well received throughout the world, thanks to better results than either the human or the machine alone. It has been observed that, for a particular type of product, the team member has additional and decisive information that improves the machine's balance sheet.
If you'd like to find out more about how the solution works and how Michelin is scaling up its innovation, you have 2 options: watch the members-only replay or ask us your questions. And if you're involved in innovation in your company, you can join the group by contacting us by e-mail.
Key success factors
- The mistake not to make: don't choose AI subjects for which you don't know how to exploit the results in the field.
- Team: its members, autonomy and ability to interact with the business
- The 6-step methodology
- The D@ril platform
- A gradual, locally-driven ramp-up