Artificial Intelligence at Roche Diagnostics

Roche Diagnostics is one of the two main divisions of parent company F. Hoffman-La Roche AG, a Swiss multinational healthcare company that focuses on pharmaceuticals and diagnostics. Roche Diagnostics’ North American headquarters is located in Indianapolis, Indiana. 

The division leverages advanced technologies to enhance its global service offerings, particularly in optimizing supply chains and sales operations through AI-driven solutions.

Roche Diagnostics reported in its annual filing that core operating profit rose by 14% and core earnings per share increased by 7%. In Roche Diagnostics, just over 46,000 people are employed globally, contributing to the more than 100,000 people that the parent company, Roche Group, employs.

Roche Diagnostics has made significant investments in AI across its diagnostics portfolio, both through integrations and collaborations. The subsidiary has a broader commitment to harnessing AI and data-driven platforms to improve diagnostic accuracy, accelerate decision-making in pathology, and support targeted therapeutic strategies.

This article examines two AI use cases at Roche:

• AI-driven predictive analytics to enhance supply chain service operations: Leveraging AI and machine learning (ML) to forecast demand, manage inventory, and respond to service needs for parts and equipment
• Machine-learning-powered recommendation engine to drive sales: Using machine learning embedded into the CRM environment to provide sales reps with real-time content and customer-engagement recommendations.

AI-Driven Predictive Analytics to Enhance Supply Chain Service Operations

Worldwide disruptions to supply chains pose a significant threat to global pharmaceutical and diagnostic brands, such as Roche, where the timely availability of parts and equipment is crucial for maintaining operational diagnostic systems in healthcare settings.

The pandemic in 2020 revealed the systemic vulnerabilities that existed when companies rely on just-in-time inventory systems and single-source supplier strategies. A 2022 study by Logistics Management showed that over 70% of global supply chains experienced disruptions in 2022. The adverse effects of disruption are far-reaching and include:

• Delays
• Increased costs
• Reduced service reliability

Supply chain volatility poses significant financial and operational risks, but its impact on patient care is equally critical. Delays in cancer treatment — often linked to equipment or supply disruptions — can increase mortality risk, as shown in a 2020 meta-analysis published in The BMJ (source). Minimizing downtime for diagnostic equipment is essential to ensure timely diagnoses and uninterrupted care delivery.

Roche turned to AI to avoid and mitigate these issues. They partnered with Baxter Planning to address these issues by leveraging AI to accomplish related goals, including:

• Predicting demand fluctuations
• Minimizing stockouts
• Dynamically managing inventory across its global network

The AI-powered platform leverages multiple data sources to drive predictive analytics and decision-making. The data sources include:

• Historical usage patterns of spare parts from past service events
• Real-time service requests and failure reports from diagnostic equipment in the field
• Inventory levels both in Roche’s global warehouses and third-party logistics partners
• Variety of other information, including equipment deployment trends, regional demand fluctuations, and supplier delays

The above data is integrated with Roche’s existing systems, including ERP and CRM. The result is a unified data flow that utilizes both supervised and unsupervised models to analyze patterns, generate demand forecasts, and simulate scenarios.

Following the initial roll-out announcements, Roche has not published any follow-up case studies or resulting outcomes from their deployment of BaxterPredict. However, Baxter Planning has published case studies that show the benefits some of their other partners have achieved. 

One case study shows a 50% decrease in inventory dollars per contract, along with a 99% boost in service levels for Palo Alto Networks.

Machine Learning-Powered Recommendation Engine to Drive Sales

Mediocre customer experiences and inefficient outreach can be costly for life sciences companies. Addressing these inefficiencies is crucial to allocating resources effectively and avoiding unnecessary costs, ultimately increasing returns. 

Recent research from the American Health Information Management Association (AHIMA) shows that predictive analytics enables healthcare organizations to shift from reactive to proactive operations, using real-time data to guide decisions and streamline delivery models. Similarly, a 2024 report from the Society of Actuaries (SOA) outlines a reusable predictive analytics framework that helps organizations forecast outcomes and optimize resource allocation, leading to measurable improvements in performance.

Roche aimed to deliver the most relevant scientific and product information to each healthcare professional at the right time, while minimizing the sales teams’ manual effort and guesswork.. 

However, before implementing AI to achieve this, Roche first had to modernize their data platform, transforming their commercial and analytics ecosystem. They used dbt Labs’ platform to accomplish this transformation. Roche needed to be able to synthesize data from multiple internal systems with external data siloed across more than 80 countries.

In the video above, João Antunes explains the architecture and runway that enabled Roche’s recommendation engine for sales reps.

Before this shift, Roche sales and medical liaison teams had to navigate a fragmented data environment, with information spread across local CRMs, data warehouses, and analytics tools. According to dbt Labs, Roche’s data had “sprawled into a disconnected ecosystem,” with each country managing its own pipelines and business logic. Sales reps often relied on intuition rather than unified, data-driven insights. The result was:

• Inconsistent customer experiences
• Inefficient outreach
• Limited visibility into what message actually worked

Now, Roche can integrate both internal and external data at scale, enabling the surface of insights that were not immediately apparent before. By combining CRM activity with physicians’ clinical trial participation, publication history, and social media engagement, it’s possible to identify physicians who are emerging thought leaders, creating a solid basis for tailored outreach. 

Roche sales representatives now receive AI-powered recommendations directly within the CRM system, which they can share with a specific physician based on their prior interactions.

Roche has not published quantifiable data indicating the tangible effects of their recommendation engine. However, the modernization efforts that preceded the recommendation engine required the decommissioning of four legacy platforms, resulting in a 70% reduction in costs.