AI is increasingly being used to solve clinical problems across the NHS, with significant investment and increased optimism moving healthcare’s adoption of AI from theory to reality. 

However, the process of deploying Artificial Intelligence Solutions into clinical practice is still an unknown for many clinicians and operations teams, with uncertainties not only in what solutions are available but how to implement solutions, how to prepare teams and systems, adapt workflows and manage the change and implementation processes associated with the adoption of any new solution. 

This webinar will invite clinicians who have led the deployment and adoption of AI tools in two different NHS organisations and specialities to share their experiences, and offer practical insights into the process of bringing AI into a complex healthcare environment.

Key themes:

• How to define and prioritise problems and establish desired outcomes for AI adoption.
• How to assess the suitability of AI tools to address critical clinical challenges with improved accuracy and 24/7 availability.
• how to assemble a multidisciplinary team to support effective implementation
• How to engage and train staff, and ensure appropriate change management
• Understanding the process of ongoing evaluation of AI tools.

Experts:

• Dr. Anmol Arora, Academic fellow, AI & ML Research & Medical Doctor, University of Cambridge
• Sarah Blake, Cardiology Registrar in Adult Congenital Hear Disease, Royal Brompton and Harefield Hospitals

Summary of the webinar:

This webinar, hosted by Helen Surana, Deputy Editor of BMJ Innovations, explored the implementation of AI in the NHS, featuring insights from two key speakers:

Dr. Sarah Blake, a cardiology registrar at the Royal Brompton and Harefield Hospitals, shared her experiences deploying AI in radiology.
Dr. Anmol Arora, an academic clinical fellow at UCL, discussed broader challenges of AI adoption, including regulatory barriers and data quality.

Key Learnings
Implementing AI in the NHS: A Step-by-Step Approach (Dr. Sarah Blake)
Identifying the Problem: The need for faster and more accurate chest X-ray reporting in emergency departments.

Choosing a Solution: Working with Cure.ai, an AI-driven radiology software company, which had existing approvals and validation in the UK.
Validating the AI Model: A local validation study ensured accuracy and safety before deployment.

Implementation Challenges:

• Educating clinicians on AI to reduce scepticism.
• Integrating AI into existing IT systems.
• Ensuring ongoing monitoring and evaluation through audits and discrepancy meetings.

Impact:

• Improved efficiency in chest X-ray reporting.
• Potential for automating normal scan identification, allowing radiologists to focus on abnormalities.

AI in Healthcare: Challenges and Future Directions (Dr. Anmol Arora)

Key Barriers to AI Adoption:

• Data Quality: AI is only as good as the data it learns from; NHS data is vast but often inconsistent.
• Algorithmic Bias: Ensuring AI models are trained on diverse patient data to avoid underrepresentation of minority groups.
• Regulatory Approvals: AI tools must comply with MHRA and NHS guidelines, which can slow adoption.

Privacy & Ethics in AI:

• Synthetic Data: AI-generated (but non-real) patient data could be used for training while preserving privacy.
• Federated Learning: A method allowing AI training on patient data across hospitals without sharing raw patient records.

AI in Clinical Training:

• AI can enhance learning by curating rare or complex cases for trainees.
• AI-assisted diagnostic tools should not replace clinicians but support them.

Key Takeaways

• AI implementation must start with a clear problem – selecting the right use case and validating AI performance is crucial.
• AI tools should integrate smoothly into existing clinical workflows, ensuring clinician engagement and trust.
• Data governance is essential – NHS must balance data sharing for AI training while ensuring privacy and security.
• Regulatory and ethical considerations must be addressed early, including MHRA approvals, post-market surveillance, and risk mitigation strategies.

 

• Collaboration is key – AI projects succeed when clinicians, software engineers, and regulators work together.
• Synthetic and federated data could revolutionize AI training while maintaining patient privacy.
• AI in education can enhance clinician learning by focusing on complex cases rather than replacing fundamental training.

Next Steps in AI & Healthcare

• AI-driven risk stratification models for outpatient monitoring (e.g., cardiology).
• More privacy-enhancing technologies, such as synthetic data for safer AI training.
• Increased adoption of AI in national quality improvement programs.
• Strengthening collaboration between the NHS, academia, and industry to streamline AI innovation.

This webinar provided practical insights into real-world AI implementation while addressing key challenges in regulation, ethics, and clinician engagement.

Sign up to our next webinar here.