Few figures have shaped the UK healthcare AI landscape like Haris Shuaib. His trajectory from medical physicist to founding the UK's first AI team within the historic walls of Guy's and St. Thomas's Hospital, to now leading Newton's Tree's cutting-edge work in scaling AI deployment in healthcare, represents an exceptional convergence of clinical depth, technical expertise, and entrepreneurial innovation.

In this episode, he brings that rare combination of perspectives to bear, offering one of the most insightful and technically informed explorations of healthcare AI's present challenges and future potential.The conversation explores how healthcare AI has evolved from early radiology-focused algorithms to today's operational workflow tools that address the daily challenges of healthcare delivery.

Shuaib discusses the critical but often overlooked challenge of post-market surveillance,  revealing how AI systems can fail in unexpected ways due to seemingly minor changes in hardware, protocols, or human behavior. He also shares eye-opening examples of automation bias, where clinicians gradually over-rely on AI recommendations, and data quality issues that can dramatically alter AI performance without any visible changes to human observers.

Chapter Guide:

01:05  Introduction

05:59 The Three Phases of Medical AI

A journey through healthcare AI's evolution from academic curiosities to the rise of "homebrew AI" that's changing who builds healthcare solutions.

08:57 The Three Ingredients Problem

Why traditional approaches to building healthcare AI might be fundamentally flawed, and the economics of where AI development should actually happen.

12:08 Automation Bias

How AI is creating new problems for healthcare professionals by filtering out "easy" cases and forcing clinicians to work at their cognitive limits.

14:56 The Hidden Consumer

Why human intuition fails when it comes to predicting AI behavior, with striking examples of imperceptible changes that completely flip diagnoses.

18:07 What Happens When We Actually Measure

Real-world evidence from monitoring systems reveals surprising patterns in human-AI interaction and the unexpected impact of routine maintenance.

20:53 How AI Amplifies Healthcare Bias

Why healthcare AI doesn't just inherit existing biases but systematically makes them worse, with concerning implications for patient equity.

23:56 Behavior Over Statistics

Why traditional AI monitoring approaches miss the point, and a new strategy for detecting problems before they cause harm.

32:33 The Power of a Single Word

How changing one term in medical documentation can flip AI predictions, revealing the subtle ways physician bias gets encoded into algorithms.

35:24 Missing Patients and Biased Language

Deeper exploration of bias manifestation, from absent demographics in training data to the amplification of clinical prejudices.

38:14 From Witchcraft to Information Science

Shuaib's vision of medicine's inevitable transformation and why future doctors will need entirely different skill sets than today.

42:32 Building the Invisible Intelligence Layer

The technology stack vision that makes AI seamless infrastructure, inspired by the most successful medical innovation of the past 50 years.

45:58 Economics of AI Adoption

A revealing comparison exposing how healthcare payment models fundamentally shape AI implementation strategies across different countries.

50:25 The Evolving Misconceptions

How assumptions about AI are changing as more healthcare professionals get hands-on experience, and why some old rules no longer apply.

56:37 Monitoring the Next Generation

Newton's Tree's ambitious project to monitor large language models in healthcare.

01:02:46 Making AI Boring

Shuaib's ultimate vision.