🤖 AI in the NHS Newsletter #9
Issue #9 | 26th June - 2nd August 2025
📋 Executive Summary
This week witnessed intense debates around AI's role in neurodivergence diagnosis, heated discussions about TPP's profit margins, and growing concerns about data governance in AI tools. The group grappled with fundamental questions about AI alignment whilst celebrating technical breakthroughs and sharing practical implementation insights. A particularly vibrant period that showcased the community's ability to balance optimism with healthy scepticism.
🎯 Major Topic Sections
🧠 The Great ADHD Diagnosis Debate
The group was divided over a research paper proposing AI for ADHD diagnosis, sparking one of the most substantive discussions of the period. The clinical safety expert raised critical concerns about AI's fundamental training to "please people," arguing that when combined with subjective diagnostic criteria, "nearly everyone will end up with a positive diagnosis."
The GP specialist highlighted practical challenges, noting that "diagnosing without prescribing is essentially an absolute non-starter" and explained the complex referral pathways that make AI diagnosis problematic. Meanwhile, the healthcare innovation catalyst suggested there might be value in "identifying those who may otherwise not present/seek confirmation" through targeted data analysis.
A fascinating counterpoint emerged from the surgical oncologist, who proposed focusing on management rather than diagnosis, referencing NHS-promoted self-management apps. The discussion revealed deep tensions between technological capability and clinical responsibility, with participants acknowledging both the desperate need created by waiting lists and the risks of oversimplified solutions.
Key Quote: "The current screening questionnaire essentially asks 'Do you think you have ADHD?' - not everything is AI-able."
💰 TPP Profit Margins: The £40m Question
A significant discussion erupted around TPP's financial performance, with revelations that the company made £40m profit on £80m revenue - a staggering 50% margin. The procurement specialist noted that NHSE had previously asked whether TPP could be disqualified from bids based on their owner's controversial remarks, but "unfortunately it was allowed to bid & they won the contract as well."
The technical infrastructure analyst advocated for open book contracts for core systems, arguing that "the NHS doesn't procure centrally so there is little/less volume based pricing." The discussion highlighted fundamental problems with NHS procurement, with participants expressing frustration about "weak procurement teams who fear legal challenges."
Key Insight: The group consensus pointed towards the need for central ownership and cultural shifts in procurement, with calls for fair pricing on integrations and sustainable deals for all parties.
🔐 Data Governance & Consent Concerns
AI scribe consent emerged as a practical battleground, with the GP practitioner asking whether implied consent sufficed given clinical settings. The information governance specialist emphasised explicit consent requirements, particularly because "audio is being sent to a third party company/software which has significant commercial interests in data handling."
The clinical implementation lead shared North East London's approach, implementing Article 6(1)(e) and Article 9(2)(h) as legal basis whilst "informing patients and providing the option to decline." The discussion revealed the complexity of balancing practical implementation with regulatory compliance.
Memorable Exchange: "I would like to know if my voice is sent to Palantir" - highlighting legitimate concerns about data destinations.
🛠️ Technical Infrastructure Evolution
The group engaged with cutting-edge developments, from Anthropic's personality vectors research to MLPerf Client 1.0 release. The AI research specialist shared excitement about Qwen models being "on par with Gemini & miles better than LLaMA4," whilst discussions around NVIDIA's DGX Spark workstation revealed appetite for on-premise AI deployment.
Edge AI engineering dominated technical conversations, with the systems engineer sharing insights about constraints when "you can't rely on just throwing more CPU/GPU grunt at it." The community demonstrated sophisticated understanding of the economic and technical trade-offs in AI deployment.
📊 Enhanced Statistics Section
Activity Metrics
Total Messages: 200+ over the analysis period
Peak Activity Days: July 30th-31st, August 1st
Average Messages per Active Day: 25-30
Busiest Time Periods: Morning (07:00-10:00) and Evening (19:00-22:00)
Top 5 Contributors with Characterising Titles
Clinical Safety Expert - 18 messages (Technical implementation & regulatory focus)
GP Innovation Specialist - 15 messages (Practical clinical applications)
Technical Infrastructure Analyst - 14 messages (Systems architecture & procurement)
AI Research Specialist - 12 messages (Cutting-edge developments & academic insights)
Healthcare Innovation Catalyst - 11 messages (Strategic thinking & industry connections)
Hottest Debate Topics by Engagement
AI for ADHD Diagnosis - 25+ messages, 8 participants, 3-hour active discussion
TPP Profit Margins & Procurement - 20+ messages, 6 participants, spanning 2 days
AI Scribe Consent Requirements - 18+ messages, 7 participants, practical focus
Data Governance for AVT Tools - 15+ messages, 5 participants, regulatory emphasis
Open Source vs Commercial AI Models - 12+ messages, 4 participants, cost analysis
Discussion Quality Metrics
Evidence-Based vs Opinion Ratio: 70:30 (High quality maintained)
Cross-Expertise Representation: 8+ different professional backgrounds contributing
Average Thread Length: 12 messages per major topic
Resolution Rate: 60% of discussions reached practical conclusions
External Resource Sharing: 15+ academic papers, regulatory documents, and technical resources
Engagement Patterns
Response Rate to Technical Questions: 85%
Follow-up Implementation Discussions: 40% of topics generated practical next steps
Constructive Debate Measures: 95% remained collegial despite contentious topics
International Perspective: Growing participation from colleagues outside England
😄 Lighter Moments
The group's humour remained delightfully intact despite serious discussions. The technical humorist perfectly captured the mainstream adoption moment: "Apple have been spamming the adverts in the cricket today with Apple Intelligence. The advert is about cat photos. We've reached the mainstream cat photo stage of the tech."
A particularly amusing exchange emerged around VPN usage when someone's foreign IP address was exposed, leading to jokes about "exposing use of VPN set to 'forrin'" and references to political figures "moaning about VPNs."
The community's self-awareness shone through discussions about being "formerly known as geeks may now be labelled neurodivergent," whilst technical troubleshooting reached philosophical heights with the ancient wisdom: "Man build house from potato. Man is idiot. Profound...."
The ongoing inside joke about one member's supposed appearance in "the Netflix Dallas Cowboys show" continued to generate chuckles, alongside playful references to "thought crime" reports to the GMC.
💬 Quote Wall
"The problem is that human engineers try to influence it despite not really knowing how it works. Probably like the first surgeons going 'I wonder if I can make their mental health better by drilling holes in the front of their head?'" - Technical Infrastructure Analyst
"I personally think we will shift to a different paradigm where we assess more individual skillsets and match them to workflows and opportunities rather than the current approach of 'You are dyslexic, here's 15 more minutes per hour per exam'" - Healthcare Innovation Specialist
"We've reached the mainstream cat photo stage of the tech. Next it'll be counter-culture rejected by the cool kids who insist only Photoshop altered photos are the real ones." - Technical Humorist
"I am amazed at what they can do. I don't really understand how they do it which perhaps makes it even more amazing" - Clinical Implementation Lead
"The biggest reason for needing consent every time is that the audio is being sent to a third party company/software which has significant commercial interests in data handling." - Information Governance Specialist
"Weak procurement is why we still gnash our teeth in anguish. So much could be achieved with what we have already but we are shackled with huge chains" - Clinical Transformation Advocate
"Just put on the reason that you want to learn from best practices across the NHS" - Implementation Evangelist (on accessing shared resources)
"I think we need to focus on management. There are several apps already promoted by the NHS for self management of more benign ends of MH" - Surgical Oncologist
🔮 Looking Ahead
Upcoming Events & Opportunities
State of AI Survey: Community encouraged to participate in Nathan's comprehensive AI adoption and sentiment research
Digital Transformation Summit: Birmingham event scheduled for September with livestream options
X-On Health Webinar Series: 15-minute lunchtime sessions running through August for Phonebar optimisation
Ongoing Themes to Watch
AI Alignment Research: Growing interest in Anthropic's safety research and its healthcare applications
Regulatory Framework Development: Continued evolution of consent and governance practices
On-Premise AI Deployment: Technical discussions likely to intensify around local model hosting
NHS Procurement Reform: Sustained pressure for central procurement and fair pricing models
Unresolved Discussions
Standardised Consent Processes: Need for sector-wide guidance on AI tool consent
Clinical Effectiveness Evidence: Calls for more robust outcome measurement from AI implementations
Neurodivergence Pathway Redesign: Fundamental questions about AI's role in mental health diagnosis
Open Source Healthcare AI: Growing interest in collaborative development models
Technical Developments
Edge AI Engineering: Continued exploration of resource-constrained AI applications
Multi-Modal AI Tools: Discussions around vision and audio processing capabilities
Interoperability Standards: Ongoing need for better system integration approaches
🌟 Group Personality Snapshot
This community continues to exemplify the best of professional discourse in healthcare technology. The group maintains its remarkable ability to navigate contentious topics with respect whilst never shying away from difficult questions. Members demonstrate genuine intellectual humility, acknowledging when they don't understand complex systems whilst still engaging thoughtfully with the implications.
The community's strength lies in its diversity of expertise - from clinical safety to procurement, from AI research to practical implementation. This creates a natural system of checks and balances where technical enthusiasm is tempered by clinical reality, and regulatory caution is balanced by innovation imperative.
What makes this group special is its commitment to evidence-based discussion. Members consistently share research, challenge assumptions, and admit uncertainties. The humour serves not just as relief but as a bonding mechanism that allows serious professionals to remain human whilst grappling with profound changes in healthcare.
The group has evolved into a rare space where the future of healthcare technology is being thoughtfully constructed through daily discourse, combining the rigor of academic debate with the urgency of practical implementation. It's a community that neither blindly embraces nor reflexively rejects change, but rather seeks to understand and shape it responsibly.
📝 Editor's Note
This newsletter captures a pivotal moment in healthcare AI adoption. The discussions reflect a maturing understanding of both the potential and limitations of AI in clinical settings. The community's growing sophistication in balancing innovation with safety concerns suggests we're moving beyond the early hype phase into more nuanced implementation strategies.
The ongoing tension between technological capability and regulatory responsibility will likely define the next phase of healthcare AI development. This group's thoughtful approach to these challenges positions it as an influential voice in shaping responsible AI adoption across the NHS.
Next Issue Preview: Watch for developments in AI alignment research applications, updates on NHS procurement reforms, and continued evolution of clinical AI governance frameworks.
Want to contribute to future newsletters? Share interesting research, pose thoughtful questions, or suggest topics for deeper exploration. This digest thrives on the community's collective intelligence and diverse perspectives.
Issue compiled with analysis of 200+ messages from June 26th - August 2nd, 2025. All participant identities anonymised whilst preserving the essence of their contributions.
