Dataiku: spotlight on Rabobank’s big data and AI strategy

In the past year and a half, Rabobank has completed more than 100 AI projects and has reduced the time to onboard data team members — in particular data scientists — from months to weeks.

The bank’s ability to do so comes out of their approach to: organizational structure; tackling a wide range of use cases; creating an innovation funnel for use cases; the education and upskilling of staff; and technology.

When it comes to raising data maturity across an entire organization, there are three main takeaways:

  • It’s not just about technology — the bigger challenge is organizing everything else around it. While the right technology can certainly make the journey smoother and be a catalyst for change, it’s not a magic bullet that will raise an organization’s maturity around data, machine learning, or AI in and of itself.
  • It’s all about the business — from selecting the right data projects to delivering value, the business is front and center. Running data initiatives from an IT team or even a siloed data team alone can never bring the depth and value brought by domain experts. The secret is getting the two to collaborate deeply together.
  • It’s a constantly evolving journey — the processes and approaches Rabobank created in 2014 are no longer applicable, and the team has adjusted them accordingly. In five years as they become even more mature, systems will continue to shift to fit their growing needs. For companies just getting started, Rabobank wouldn’t recommend beginning with their current approach — it’s something you need to grow toward, customizing for particularities of your organization.

Machine learning operations (MLOps) moves to forefront

One of the areas that has emerged in the last year as an important pillar of Rabobank’s “Analytics Way of Working” is model maintenance. When the team first introduced the Innovation Funnel, the focus was on generating new ideas. As the entire organization has advanced its data maturity, attention has shifted to ensuring models stay updated, that there is a good audit trail for models already in production (especially in the KYC domain), strong data lineage and quality, etc.

According to Roel Dirks, product manager in Rabobank’s Big Data Lab, global economic changes amid the COVID-19 pandemic put the Analytics Way of Working into the spotlight — models that were created around the organization outside of this framework were taking up to several years to develop and validate. In 2020 and 2021, everyone needed to be more agile, so taking years to develop one model or dashboard was no longer an option.

“We cannot think about model maintenance when we need maintenance; we need to think about it in the development phase,” Dirks said in the report.

Read the full report

Related Posts

Previous Post
SFM Interview: Capstone’s treasurer Lupea on regulation, market disruption and ESG
Next Post
FinCEN identifies ransomware trends in BSA data for H1

Related Posts

Fill out this field
Fill out this field
Please enter a valid email address.


Reset password

Create an account