How Relevant Software Built an AI-powered Analytics Platform for AstraZeneca
Medical Affairs teams rarely struggle with data scarcity. They struggle with data that arrives in the wrong shape for decisions. Field notes sit in free text, clinical documentation lives in long PDFs, real-world evidence comes with inconsistent structure, and CRM entries capture nuance that never makes it into reports. The result is often the same across pharma: slow insight cycles, duplicated effort, and an uncomfortable trade-off between speed and control.
A short case we would like to offer is a concrete example of how teams can break that pattern. The core story centers on AstraZeneca and how the delivery team at Relevant Software built a compliant AI workflow that converts scattered Medical Affairs inputs into structured outputs teams can use quickly.
The bottlenecks that show up in almost every pharma Medical Affairs workflow
These problems repeat because Medical Affairs work depends on language, context, and traceability, not only on raw numbers.
Teams typically hit four bottlenecks:
- Unstructured evidence at scale. Field insights and medical context arrive as text; someone must interpret, summarize, and map them to a reporting format.
- Manual CRM processing. People copy, check, and reconcile entries across sources, which slows cycles and creates avoidable errors.
- Hard compliance boundaries. Sensitive data requires encryption, strict access rules, and clear governance. Many «fast» approaches fall apart here.
- Slow knowledge reuse. Insights stay trapped in individual notes and local files, so teams repeat the same synthesis work week after week.
AstraZeneca faced the same dynamics. Medical Affairs teams handled a growing volume of complex, unstructured medical data, including clinical trial protocols, patient observations, and notes from conversations with healthcare professionals and manual processing delayed access to critical insights.
What AstraZeneca needed, stated in practical terms
The project started with a direct question: how to automate CRM record analysis without compromising accuracy or compliance.
The solution had to process three data streams that behave differently in production:
- Clinical trial documentation, such as protocols, treatment results, and patient safety observations.
- Real-world evidence, such as doctor feedback, patient stories, and healthcare reports.
- Field notes recorded by Medical Science Liaisons during conversations with healthcare providers.
The system also had to remove processing bottlenecks, deliver quicker access to key information, and enable teams to respond faster to research and market needs.
How Relevant Software solved the problem
Relevant Software joined the project as the delivery partner to design an AI solution that turns raw, fragmented Medical Affairs data into structured, compliant, ready-to-use reports. The implementation focused on four pillars that map to how Medical Affairs teams operate in practice.
Automated insight extraction from medical text
The system scans large volumes of field notes, research data, and clinical reports, then extracts key insights to support reporting and decision-making while maintaining compliance requirements.
This matters because it shifts work from manual processing to review and judgment, which is where Medical Affairs teams add the most value.
Real-time processing on Google Cloud
The solution runs on Google Cloud to provide fast, stable access and to support large teams, global operations, and increasing volumes of medical information.
Real-time access also appeared in the explicit client requirements, along with high-accuracy insight extraction and full regulatory compliance.
Security and regulatory compliance are built into the architecture
The system protects sensitive data with encryption and role-based access controls, so each user sees only the information required for their role. This approach reduces the risk of data leaks, helps prevent human error, and aligns with healthcare regulations, including GDPR and HIPAA.
Instead of treating governance as paperwork at the end, the project embedded it into daily usage patterns.
Scale across teams without losing speed or accuracy
Medical Affairs adoption tends to expand quickly once teams trust the outputs. The case highlights scalability as a core capability, with the AI solution handling large datasets across teams and departments while maintaining performance.
On the engineering side, the delivery included building the system using ChatGPT and Llama 2, fine-tuning the models for Medical Affairs interpretation, deploying it on Google Cloud, and integrating encryption, access controls, and regulatory safeguards.
What changed after delivery
The results read like operational improvements that a Medical Affairs leader can defend with numbers.
- 25% efficiency increase for Medical Science Liaisons and 20 hours per week saved on manual CRM data processing
- 35% reduction in market access budget by eliminating redundant processes
- 40% faster preparation of medical materials, driven by automation of routine data search and document preparation
- 35,000+ records processed automatically, with collection, checks, and structuring across sources
- 25% fewer data errors, which improved confidence in reports and presentations prepared for doctors and regulators
The client feedback reinforces the same theme. A Director in bio-pharma medical evidence described the outcome as a «precise and compliant AI system» and credited NLP plus secure data management as essential to success.
A simple takeaway teams can reuse
When Medical Affairs analytics relies on manual interpretation of text, speed and consistency collapse first, then quality follows. This case shows a workable alternative: treat insight extraction, access control, and compliance as a single connected engineering problem, then build a system that produces structured outputs quickly, at scale, with governance that holds up under real-world usage.
How Relevant Software Built an AI-powered Analytics Platform for AstraZeneca
Medical Affairs teams rarely struggle with data scarcity. They struggle with data that arrives in the wrong shape for decisions. Field notes sit in free text, clinical documentation lives in long PDFs, real-world evidence comes with inconsistent structure, and CRM entries capture nuance that never makes it into reports. The result is often the same across pharma: slow insight cycles, duplicated effort, and an uncomfortable trade-off between speed and control.
A short case we would like to offer is a concrete example of how teams can break that pattern. The core story centers on AstraZeneca and how the delivery team at Relevant Software built a compliant AI workflow that converts scattered Medical Affairs inputs into structured outputs teams can use quickly.
The bottlenecks that show up in almost every pharma Medical Affairs workflow
These problems repeat because Medical Affairs work depends on language, context, and traceability, not only on raw numbers.
Teams typically hit four bottlenecks:
- Unstructured evidence at scale. Field insights and medical context arrive as text; someone must interpret, summarize, and map them to a reporting format.
- Manual CRM processing. People copy, check, and reconcile entries across sources, which slows cycles and creates avoidable errors.
- Hard compliance boundaries. Sensitive data requires encryption, strict access rules, and clear governance. Many «fast» approaches fall apart here.
- Slow knowledge reuse. Insights stay trapped in individual notes and local files, so teams repeat the same synthesis work week after week.
AstraZeneca faced the same dynamics. Medical Affairs teams handled a growing volume of complex, unstructured medical data, including clinical trial protocols, patient observations, and notes from conversations with healthcare professionals and manual processing delayed access to critical insights.
What AstraZeneca needed, stated in practical terms
The project started with a direct question: how to automate CRM record analysis without compromising accuracy or compliance.
The solution had to process three data streams that behave differently in production:
- Clinical trial documentation, such as protocols, treatment results, and patient safety observations.
- Real-world evidence, such as doctor feedback, patient stories, and healthcare reports.
- Field notes recorded by Medical Science Liaisons during conversations with healthcare providers.
The system also had to remove processing bottlenecks, deliver quicker access to key information, and enable teams to respond faster to research and market needs.
How Relevant Software solved the problem
Relevant Software joined the project as the delivery partner to design an AI solution that turns raw, fragmented Medical Affairs data into structured, compliant, ready-to-use reports. The implementation focused on four pillars that map to how Medical Affairs teams operate in practice.
Automated insight extraction from medical text
The system scans large volumes of field notes, research data, and clinical reports, then extracts key insights to support reporting and decision-making while maintaining compliance requirements.
This matters because it shifts work from manual processing to review and judgment, which is where Medical Affairs teams add the most value.
Real-time processing on Google Cloud
The solution runs on Google Cloud to provide fast, stable access and to support large teams, global operations, and increasing volumes of medical information.
Real-time access also appeared in the explicit client requirements, along with high-accuracy insight extraction and full regulatory compliance.
Security and regulatory compliance are built into the architecture
The system protects sensitive data with encryption and role-based access controls, so each user sees only the information required for their role. This approach reduces the risk of data leaks, helps prevent human error, and aligns with healthcare regulations, including GDPR and HIPAA.
Instead of treating governance as paperwork at the end, the project embedded it into daily usage patterns.
Scale across teams without losing speed or accuracy
Medical Affairs adoption tends to expand quickly once teams trust the outputs. The case highlights scalability as a core capability, with the AI solution handling large datasets across teams and departments while maintaining performance.
On the engineering side, the delivery included building the system using ChatGPT and Llama 2, fine-tuning the models for Medical Affairs interpretation, deploying it on Google Cloud, and integrating encryption, access controls, and regulatory safeguards.
What changed after delivery
The results read like operational improvements that a Medical Affairs leader can defend with numbers.
- 25% efficiency increase for Medical Science Liaisons and 20 hours per week saved on manual CRM data processing
- 35% reduction in market access budget by eliminating redundant processes
- 40% faster preparation of medical materials, driven by automation of routine data search and document preparation
- 35,000+ records processed automatically, with collection, checks, and structuring across sources
- 25% fewer data errors, which improved confidence in reports and presentations prepared for doctors and regulators
The client feedback reinforces the same theme. A Director in bio-pharma medical evidence described the outcome as a «precise and compliant AI system» and credited NLP plus secure data management as essential to success.
A simple takeaway teams can reuse
When Medical Affairs analytics relies on manual interpretation of text, speed and consistency collapse first, then quality follows. This case shows a workable alternative: treat insight extraction, access control, and compliance as a single connected engineering problem, then build a system that produces structured outputs quickly, at scale, with governance that holds up under real-world usage.
