Michelin: From Siloed Data to a Global AI Fabric With Dataiku

Even for a company with Michelin’s global footprint (121 production sites, 132,000 employees, and a reputation for engineering excellence) data wasn’t always easy to access, share, or apply. Engineers and technicians often had to rely on coding experts to access relevant datasets, models weren’t reusable across teams or factories, and discovering insights could take months of manual work.

The Problem: Diversified Workflows 

Historically, AI efforts at Michelin were diversified. Technical teams and coding experts controlled access to data and models, while business experts like engineers and technicians had limited direct involvement. Reuse across sites was rare, insights were slow to emerge, and value creation was uneven. This recurring problem slowed down innovation and left production teams without timely data to optimize operations.

As Michelin’s data landscape expanded in both volume and complexity, traditional methods proved insufficient. With a vast array of applications, distributed IT systems, and substantial data growth, the organization recognized the imperative for scalable AI, not merely to enhance operational efficiency, but to accelerate innovation in tires and mobility solutions and to advance its long-term sustainability objectives.

A Unified Data Foundation With Dataiku

To overcome these challenges, Michelin sought a platform that could work across cloud, on-prem, and hybrid environments while serving both technical and non-technical users. Just as importantly, the solution needed to:

• Enable reuse of use cases across sites, multiplying impact
• Support both code-based and visual workflows
• Connect seamlessly with Michelin’s broad tech ecosystem, including Snowflake, Databricks, Azure, Kafka, and Power BI

Dataiku, The Universal AI Platform™, was selected in 2021 as the foundation for this strategy, starting with a small group of 35 users in digital manufacturing. By mid-2025, adoption had expanded to 1,500+ users across 50+ factories, with 80% being business experts such as process engineers, quality technicians, and R&D specialists. Michelin’s goal is to soon deploy Dataiku across all factories.

Dataiku is already a key enabler of our Industry 4.0 transformation, enabling daily usage of AI at the heart of our factories, driving sustainable performance. Matthieu Leynet Manager, Innovation for Industry 4.0, Michelin

Deployment and Adoption Across Michelin

Dataiku is now deployed across Michelin’s manufacturing operations, corporate functions, and R&D. Seventy percent of users are based in factories, where Dataiku supports predictive maintenance, quality control, energy optimization, and process improvement. Others use the platform for sustainability programs, new mobility services, and tire innovation, reflecting the breadth of use cases made possible with Dataiku.

This adoption has been enabled by Dataiku’s:

• Hybrid deployment model, supporting both on-premise installations in sensitive R&D centers and cloud-based environments connected to Azure, Snowflake, and Databricks.
• Interoperability across systems, including deep integration with the AVEVA PI system deployed in more than 65 factories.
• Collaborative architecture, allowing internal teams and external partners to co-develop and share projects within governed instances.
• Reusable project templates and automation interfaces, enabling teams to replicate successful solutions across sites while maintaining consistent data and model standards.

As AI adoption accelerated across factories and R&D, Michelin faced a familiar challenge: scaling innovation without losing oversight or consistency. Michelin uses Dataiku’s governance to balance global oversight and local innovation, forming the backbone of its AI ecosystem.

Operational Use Cases That Deliver Tangible Gains

Michelin teams now use Dataiku to solve problems faster, more consistently, and with greater impact. 

Transforming Quality Management at Scale

More than 600 process engineers and quality technicians across 10+ factories rely on the Dataiku Parameters Analyzer Solution to identify the root causes of recurring defects on production lines. While the influencing factors (workshop temperature, product age, machine type) were intuitively known, their complex interrelationships had never been modeled. Instead of sifting through thousands of production variables manually, a process that used to take up to six months, they can now run analyses in about an hour. 

The tool not only ranks the variables most correlated with inefficiencies, but also lets engineers test improvement scenarios and share results instantly with technicians and data scientists, accelerating decision-making across the factory floor. 

Thanks to Dataiku, we can put the use of data in the hands of technicians in charge of quality, maintenance, and process improvement.” Marine De Cezac Director Industry 4.0, Michelin

Mobility Services & IoT

Michelin also uses Dataiku to process data from sensors embedded in tires and vehicles. This fuels services like real-time pressure loss alerts and end-of-life predictions, helping commercial fleets reduce costs and increase safety.

The integration between Dataiku and Snowflake allows Michelin to simplify the adoption of data and AI for both coders and non-coders and accelerate the innovation of the group. Julien Millau IT Architect, Michelin

Results: ROI & Business Value 

With Dataiku, Michelin has achieved:

• Faster insight delivery, with analyses reduced from months to hours
• Tangible operational savings
• Empowerment of 1,500+ active users, the majority of whom are business experts

Together, these gains drive both bottom-line efficiency and top-line innovation, delivering measurable ROI across factories, services, and R&D.

What’s Next: Further Integrating GenAI & Agents Into Operations

Michelin is already preparing the next stage of its AI journey: integrating GenAI and intelligent agents into daily operations. Examples include: 

• A “talk to my time series data” chatbot, using Dataiku Answers with Azure OpenAI, that allows engineers to interact directly with time series data (e.g., pressure, temperature, injection cycles). The agent can calculate statistics, detect anomalies, generate visualizations and reports, and even perform root cause analysis, all through a simple conversational interface.
• An agent-enabled Parameters Analyzer that lets engineers run the entire workflow through a chatbot interface with human-in-the-loop (HITL) oversight. Engineers ask questions, guide the analysis, and collaborate with the agent to identify optimal outcomes. The agent then generates instant summaries and reports, acting as a true teammate on the factory floor.
• Indexing technical documentation in R&D to unlock and leverage internal knowledge, simplify access, and prevent redundancies across approaches, protocols, and tests

Beyond these initiatives, Michelin is laying the foundation for long-term scale. With the Dataiku LLM Mesh, the company can securely deploy universal standards across all environments while flexibly tapping into multiple LLM providers and service layers, ensuring agility and consistency.