John Lewis Partnership: Ensuring MLOps at Scale
With Deloitte & Dataiku

John Lewis Partnership (JLP) faced several challenges in their journey to operationalize machine learning (ML) and AI models. Although they had a talented team of over 20 data scientists, their AI models were stuck in the experimental phase, unable to be deployed into production. This prevented JLP from scaling their AI initiatives, realizing financial benefits, and getting a full return on investment.

Partnering With Deloitte and Dataiku

The primary issue was the lack of MLOps capabilities within the data analytics team. There were no established tools, best practices, or governance frameworks for deploying production-grade models. Integrating MLOps into JLP’s existing operating model was also a challenge.

JLP partnered with Deloitte on their AI operationalization challenges, using Dataiku as the foundational platform. Deloitte identified top-tier engineering expertise as critical to solving these issues and ensuring that JLP’s AI models could transition from development to a production environment. Together, they implemented a comprehensive solution focused on establishing a solid MLOps foundation, accelerating model deployment, and empowering JLP for long-term success.

• MLOps Foundation: Deloitte conducted a maturity assessment and created a tailored MLOps operating model for JLP, including a robust model delivery process and governance framework. This provided a structured and scalable approach to managing the entire AI lifecycle within Dataiku.
• Accelerating Deployment: With Dataiku, JLP and Deloitte operationalized 12 production-grade AI models. The platform’s capabilities for model deployment, monitoring, and management — combined with pre-built AI model delivery templates — enabled a faster time-to-value for JLP’s models.
• Empowering for the Future: Deloitte provided upskilling programs for JLP’s data scientists on MLOps principles and Dataiku best practices. Additionally, service support and operational runbooks ensured smooth model maintenance and operation.

Impact on Day-to-Day Operations

The implementation of the Dataiku and Deloitte solution has revolutionized JLP’s AI strategy and daily operations. Previously, deploying models to production could take months, but now JLP can achieve it within five weeks on average. Model iteration times have also dramatically reduced from 10-12 weeks to just one to two weeks, enabling the business to respond more quickly to changing market dynamics.

The solution has fostered a more collaborative environment, where multidisciplinary teams work together seamlessly to extract maximum value from AI models. Standardized processes, robust governance, and clear operational frameworks have replaced the previously disorganized approach. Most importantly, JLP’s data science team has gained the ability to build and maintain models in-house, ensuring the long-term sustainability of their AI initiatives.

Value Generated

Deloitte’s work with JLP unlocked tens of millions of dollars in financial benefits, directly impacting the company’s bottom line. The operationalized AI models improved key business areas such as online checkout, product availability forecasting, and promotion planning, driving increased sales and reduced costs. AI-powered solutions also optimized bakery operations and returns processes, delivering significant savings through reduced waste and enhanced efficiency. Improved labor scheduling and delivery slot availability further boosted customer satisfaction.

Dataiku played a crucial role in the success of this initiative, particularly in empowering JLP’s data scientists and establishing robust governance processes. The Universal AI Platform’s intuitive visual interface and library of pre-built components allowed the data scientists to focus on refining algorithms without getting bogged down in infrastructure management. This ease of use enabled faster experimentation and reduced time-to-value.

Moreover, Dataiku’s governance features — such as model documentation, version control, and explainability tools — ensured compliance, transparency, and trust in AI decisions. This simplified technical complexity while providing the necessary oversight for responsible AI practices, allowing JLP to innovate confidently and deliver business value faster.