Your browser doesn’t support HTML5 audio
Digital leaders regularly tell us that the key to exploiting AI and other emerging technologies is a strong data foundation. One strong proponent of that approach is Thierry Martin, Head of Enterprise Data and Analytics at Toyota Motor Europe, who has helped his company prepare for the next wave of automation.
When diginomica last spoke with Martin in late 2023, his organization was beginning to use Snowflake’s cloud-based platform to give employees across the business access to a range of data products. The aim was to create an enterprise-wide data mesh and a business intelligence platform. While that program has progressed significantly, Martin and his team have made gains in other areas, particularly concerning AI:
In 2023, we were mostly focused on understanding where we wanted to go. We were laying down the foundations. Now, Snowflake is the de facto analytics platform. We have moved a lot of data into Snowflake, and this data is the raw material for the data products we use for reporting, analytics, and AI.
In fact, today, Martin describes Snowflake as the cornerstone of Toyota Europe’s data mesh. In the past two years, the organization has worked with Snowflake’s professional services team to refine its platform and ensure that compliance with regulations such as GDPR is baked into its processes and considered before its data products are deployed to production.
Martin’s team has worked with their line-of-business colleagues to integrate information into the Snowflake platform. Across all organizational areas, from design to logistics, his team has created a backlog of between 300 and 400 data projects. In January, they passed the milestone of launching 100 data products in their internal data marketplace.
These products, which have data owners and an associated set of metadata tags, are published as assets that anyone with the right credentials can access. Now, with one eye on refining the platform and another on embracing AI-enabled innovation, Matin and his team are looking to make more of Snowflake’s technology:
We are paving the road as we are moving forward. We are deciding which direction to take, the architecture of the road, and we have all the people using the road just behind us. And we are all doing all this work at the same time.
Identifying trends
While his team faces a complex program of work across these areas, Martin says that using Snowflake as a data foundation and platform for innovation is the key to success. He says this joined-up approach to architecture and emerging tech is a key lesson for other digital leaders:
We are doing data governance on Snowflake, but we are also building AI. We are building data products. We are doing data architecture. We are doing everything that a data analytics team should do.
Martin suggests that many other organizations take a more fragmented approach, where the data platform is part of IT, data analytics is part of business, and data governance resides in a separate silo. Toyota Europe’s joined-up approach to governance and AI means his team can provide a centralized and trusted data resource across all areas, whether that’s architectures, models, governance, or pipelines:
I think part of the success is to have a team that is vertically integrated. If someone calls us with a data question, we can answer. Of course, these questions are answered by different people, but it's the same team. That's our mission, and that's what makes us strong as a team.
Rather than a traditional data warehouse, Martin views Snowflake as a scalable computation engine that’s the core of his organization’s data mesh. Other key technologies in the data stack include Calibra for governance, Dataiku for collaboration, Qlik for ingestion, DBT for transformation, Monte Carlo for observability, and Sigma for analytics.
Martin says the 100 data products his team has launched cover a broad range of areas. He gives the example of products associated with vehicle warranties. Toyota Europe sells 1.2 million cars every year, generating a vast amount of connected and non-connected data. All this information comes back into Snowflake and the integrated data stack:
So, can we, for instance, analyze warranty repairs? Is there a trend in warranties across Europe? By moving data into Snowflake, we can make a daily report. That makes it possible to run Python, detect abnormalities, and run statistical testing on data. This approach is a changing point for the business.
Embracing innovation
Further innovation is afoot. Toyota Europe is already using Snowflake Intelligence, the tech company’s agent that allows users to exploit enterprise knowledge using natural language.
Martin is impressed by the technology’s performance. Instead of his team writing Python code and creating detailed instructions for an AI chatbot, line-of-business employees can describe to the Snowflake-powered agent in natural language what it should and shouldn’t do. He says this AI-enabled development allows his team to create data products much more quickly and for business users to generate insights rapidly:
What previously took several months, we could shorten that process to a few weeks, and now we are getting even faster.
One example is employees using an agent to query sales data. Martin says collating insights for popular vehicle specifications for each European country might previously have been a laborious process. With Snowflake technology, an agent can respond to a prompt and quickly collate data to identify key trends:
AI can help product planners to better understand the right specifications that we need to deliver to the customers.
Martin recognizes that agentic AI remains a work in progress. He says a clear strategy is crucial. Agents are not allowed to operate autonomously, and they work under strict governance and security guidelines. Just as Toyota Europe won’t create thousands of data products, he says the company’s AI strategy will focus on developing agents that can be re-used in other areas:
That's fundamental. And then, one by one, we work with our data scientists to ensure we have defined the correct semantic model, so we know what to expect from these agents. So, in the current vision, people will connect to one personal agent, which is like their own orchestrator, and from there, this service will connect to different agents, depending on their access rights.