Client

Global medical device company and healthcare operator, enabling millions of treatments worldwide.

Background

Our client needed to accelerate AI innovation to maintain its position as leading device manufacturer.

Together with our client’s strategy team, we developed an AI strategy, designed a company-wide data platform and assessed a portfolio of use cases. Out of 63 potential AI applications, we rapidly validated the three most promising and led one to full implementation.

Approach

• Clear strategic focus: Instead of starting with a bunch of undifferentiated use case ideas, the AI strategy was derived from an overall strategic rationale: evolving from a device manufacturer to a strategic partner for clinics, leveraging deep disease knowledge, vast data, and touchpoints along the whole patient journey for new products and services.

• Fix the plumbing: We designed a company-wide data platform, integrating various existing and new data pipelines, ensuring a robust foundation for AI use case implementation.

• Tangible impact over cool demos: For 63 potential pilots spanning Quality Management to Therapy Management, we screened for fast ROI, regulatory feasibility, and technical readiness to identify “lighthouse” use cases with the power to spark momentum.

• Build credibility and capability: We rapidly tested and refined three tangible use cases to demonstrate the potential and test the new platform early-on.

Outcome

• Delivered a scalable data platform architecture that enables seamless integration and future AI expansion.

• Defined and prioritized a portfolio of high-impact AI use cases aligned to the company’s strategic goals.

• Validated the commercial, regulatory, and technical feasibility of three top use cases.

• Implemented a flagship AI solution supporting medication dosing, improving treatment quality, and increasing value-based care earnings.

Key takeaway

The most precious resources: momentum and excitement that fuel support for building internal capacity and sustain a long-term AI journey.

Client

Multinational healthcare delivery company and medical device OEM focused on end-stage renal disease.

Background

In its own clinics, the client had successfully deployed a data-driven therapy optimization approach combining demographics, medication, and treatment history data. For more than a decade, the approach has significantly reduced mortality and treatment costs.

Approach

Building on this internal success, we conceptualized, validated, and piloted the approach as a productized service for third-party clinics.

This included developing a viable business model, implementing a harmonization layer to handle the heterogeneous EHR data input (HL7 FHIR v2), building a solution that could roll-out quickly, creating a patient benchmarking dataset and piloting the tool with a clinic for identifying actional treatment insights.

Outcome

Our pilot uncovered immediate treatment adjustments and was strongly endorsed by both the leading physician and the Medical Director Germany of the clinic chain. Building on this success, the solution will now be integrated into the broader therapy monitoring system.

Key take-away

To prevent slow back-and-forth between organizational silos, we needed to design solutions that 

• Fit into the technical landscape: SQL scripts, a Python core and an Excel “frontend” – nothing fancy, but “battle-tested” and implemented within hours at a clinic.

• Tick all boxes for legal and QM: we drafted not only the technical architecture, but also the data processing agreements, the procurement process and the QM process to ensure they all fit together.

• Create immediate value and trust at the frontline: For the first time, doctors could observe treatment trends across all their patients, enabling proactive intervention and reducing adverse events

Client

Early-stage drug discovery startup with focus on neurodegenerative diseases.

Background

Our client was prioritizing target candidates in preparation for a Phase 1 clinical trial. Although they had a clear concept of what an ideal target should look like, existing databases lacked the proprietary information required for their unique approach, and much of the relevant data was buried in vast amounts of primary literature.

Approach

Collaborating closely with the R&D team, we developed a LLM-based text mining engine to scan millions of publications and extract all available information on potential targets. Within days, we created a proprietary map of evidence that summarized findings across prioritization criteria, with supporting evidence, structured data, and direct source links for all target candidates.

Outcome

The map of evidence refined target prioritization and surfaced several new candidates that had not been considered by the R&D team, expanding the pipeline and strengthening the scientific rationale for next-stage discovery efforts

Key take-away

Next-generation text mining not only accelerates the collection of drug target data but also enables analysis at an unprecedented scale: screening millions of documents within days to reveal actionable insights that traditional methods would miss.