Case Study 1: Accelerating Oncology Insights Through AI-Driven Process Optimisation

1.     Executive Summary:

Led a digital transformation pilot integrating Large Language Models (LLMs) into the Systematic Literature Review (SLR) workflow for an oncology insights function. By evaluating and deploying advanced NLP and GenAI tools, we automated manual literature screening and data extraction, achieving up to a 60x improvement in task efficiency. This strategic pharma process optimisation enabled the insights function to expedite insight delivery, increase executive credibility, and allocate more of their capacity toward high-value portfolio strategy projects.

2.     The Client’s Challenge:

An Oncology Insights department was experiencing operational bottlenecks within its Systematic Literature Review (SLR) processes. The manual execution of designing search strategies, conducting evidence extraction, and generating insights often suffers from slow turnaround times.

These operational inefficiencies created other second-order strategic challenges:

·       Stakeholder Frustration: Senior leader stakeholders expect prompt insight delivery, but the team's slow responsiveness hindered strategic agility.

·       Diminished Strategic Influence: Reduced capacity to address increasing analytical demands limited the department’s overall strategic impact.

·       Threatened Credibility: The perceived inefficiency negatively impacted the function's influence at the executive level, potentially putting future departmental funding and support from leadership at risk.

·       Inflated Costs & Wasted Talent: Analysts were spending hours on manual data extraction, limiting their capacity to deliver high-value, strategic insights without requiring additional headcount.

The department urgently needed a solution to improve operational excellence and consolidate its strategic value to senior management.

3.     The Solution & Methodology:

To resolve these inefficiencies, I worked with a specialised team of three (myself, an Associate Director Subject Matter Expert in SLRs, and an AI/ML specialist) to scope and execute a pilot project evaluating the integration of GenAI into the insights function’s workflow. The objective was to evaluate and compare five ChatGPT-based text analysis tools to identify opportunities to use them for process improvement.

Key Methodologies & Interventions:

·       Process Mapping & Design: identifying the most labour-intensive steps suitable for AI automation.

·       Search Strategy & Query Design: Includes use of data sources like Google Scholar, Trialtrove, ClinicalTrials.gov, Medline, Embase, and PubMed.

·       Tool Evaluation: Tested specialised research LLMs tools against traditional manual workflows.

·       Risk Mitigation: Evaluated the tools for hallucination risks, identifying that while fabricated data and non-existent references occurred, they could be drastically mitigated through advanced prompt engineering and targeted human review.

4.     Results & Strategic Impact:

Our work demonstrated promising, measurable improvements in business efficiency. Our findings were presented to our oncology insights function in a comprehensive summary PowerPoint presentation report. Given the success of our work, we were also shortlisted to present again at a wider Cross-Therapy Area Insights community forum. The benefits of this work were 2 folds.

A.    Quantifiable Efficiency Gains:

The integration of GenAI yielded the following time savings:

·       Data Extraction (from publications tables, figures, and results): Time reduced from 30 to 120 minutes down to just 2 minutes per task, a 15x to 60x improvement in efficiency.

·       Data Summary (from publications abstracts, background, and discussion): Time reduced from 10 to 15 minutes down to 2 minutes, a 5x to 7x improvement in efficiency.

·       Data Inference: Time reduced from 30 minutes down to 10 minutes (a 3x improvement). Furthermore. AI tools used decreased limitations caused by the analyst’s unfamiliarity with niche topics.

B.     Strategic Outcomes & ROI:

The tangible productivity gains transformed the department's operational capabilities:

·       Consolidated Function’s Credibility and visibility with higher management: The demonstrable increase in productivity and faster insight delivery provided senior management with proof of increased efficiency.

·       Increased Departmental Capacity: members of the insights function team could spend less time doing frustrating manual data handling work, allowing them to manage increasing workloads and take on higher-value strategic initiatives without expanding headcount.

·       Accelerated Decision-Making: Expedited insight generation directly enabled quicker, more agile clinical and portfolio strategy decisions.

·       Catalyst for Digital Transformation: The success of this pilot drove the adoption of these tools by the CI department and led to the launch of several subsequent AI projects, including the development of a purpose-built LLM-assisted literature web scraping tool.