Prove It

Paul Wicks sits down with Dr. Tauseef Mehrali , GP, former VP of Medical Safety at Ada Health, and now VP of Regulatory Science at Flo Health, the world's largest femtech platform, for a deeply practical conversation about what it actually takes to build safety and regulatory rigour into digital health products at scale.

Flo has over 400 million downloads and nearly 80 million monthly active users. When something goes wrong at that scale, the harm is invisible, insidious and potentially enormous. Tauseef's job is to make sure it doesn't.
In this conversation they cover:

🛑 Why he turned down a job after reading their company bible, and why culture is a safety signal
🌡️ The Ada orange flag system: how to detect harm before it becomes a crisis
📱 Introducing medical device regulation into a 400M-user consumer app without breaking it
⚖️ Intended use vs actualised use vs off-label use, the grey zone nobody talks about in femtech
🏰 The regulatory moat and the drawbridge: why one without the other is worthless
📋 The single most important thing a founder can do on day one (it's not what you think)

Plus: what book should be on the shelf instead of Blitzscaling, and why Tauseef might just have to write it himself.

🎙️ Host: Paul Wicks | ProofStack Health
🏢 Guest: Dr. Tauseef Mehrali, VP Regulatory Science, Flo Health
🔗 Connect with Tauseef on LinkedIn
🔗 Learn more: https://www.proofstackhealth.com
📬 Subscribe to Proof Points for evidence frameworks built for digital health teams
https://proofpoints.proofstack.health

In 2015 Geoffrey Hinton said radiologists would lose their jobs to AI within five years. It's 2026 and we're still short of radiologists. Dr. Hugh Harvey has a few thoughts on that.

Paul Wicks sits down with Hugh, radiologist, ex-Babylon Health, Europe's first CE-marked deep learning mammography device, and now regulatory consultant to over 150 medical device companies, for a masterclass on why most healthcare AI companies fail, and what the ones that survive actually do differently.

In this conversation they cover:
🏥 Why AI didn't replace radiologists, and what that really tells us about healthcare AI
📋 How software gets classified as a medical device (and why start-ups always try to game it)
🚩 "Class one washing": the shortcut that makes your product less useful AND gets you caught
🤖 AI-generated fake regulatory documentation — it's happening and it's worse than you think
💰 The health economics gap: why regulatory clearance is just the entry ticket, not the prize
⚠️ Digital thalidomide: why a small error rate at scale is an entirely different category of harm
🏆 What the companies that actually make it do from day one
Plus: why there's no such thing as vibe-coding your way through medical device regulation.

🎙️ Host: Paul Wicks | ProofStack Health
🏢 Guest: Dr. Hugh Harvey, Regulatory Consultant, Hardian Health
🔗 Hardian Summit — April 29th, BMA House, London: visit Hardian's website for tickets
🔗 Learn more: https://www.proofstackhealth.com
📬 Subscribe to Proof Points for evidence frameworks built for digital health teams


00:00 — Introduction: The Magic Trick That Doesn't Work at Home
02:00 — Why Radiologists Still Have Jobs in 2026 (And What That Tells Us About All of Healthcare AI)
08:00 — From Babylon Health to Europe's First CE-Marked Deep Learning Device: Hugh's Journey
16:00 — How Software Gets Regulated as a Medical Device (Explained Simply)
26:00 — Class One Washing: The Shortcut That Backfires
36:00 — AI-Generated Fake Regulatory Documentation: A Growing Real Problem
42:00 — The Health Economics Gap: Why Regulatory Clearance Is Just the Entry Ticket
45:00 — Digital Thalidomide: Why Scale Changes Everything About Harm

Before mental health chatbots were a trend, Dr. Ali Darcy was already running randomised controlled trials on one.

In 2016, she and her collaborators at Stanford conducted what is widely regarded as the first ever formal RCT of a conversational AI for mental health.

The paper has since been cited over 2,000 times. Her creation, Woebot, became one of the most studied and recognised digital mental health tools in the world.

In this episode of Prove It, host Paul Wicks sits down with Ali for a wide-ranging, deeply honest conversation about what the science of AI mental health actually shows, and what it doesn't.

They cover the engagement crisis in digital CBT, why humour is a legitimate and underused therapeutic tool, the surprising finding that users form a meaningful bond with a fictional robot character, and the critical question the field still hasn't answered: what exactly is the active ingredient of the human therapist?

They also dig into the limits of large language models in mental health, why doubling NLP accuracy didn't improve outcomes, why "rushing in to help" is therapy's number one failure mode, and why role clarity may be the most important safety issue in AI mental health right now.

This is a conversation between two researchers who refuse to hand-wave. If you care about digital mental health done right, this one is for your library.

Topics covered:

• The 2016 Stanford RCT: design choices that still hold up nearly a decade later
• Engagement as the primary unsolved problem in mental health tech
• The therapeutic alliance paper: 37,000 users, 200x larger than the entire existing literature
• What Woebot got right by being transparently fictional rather than pretending to be human
• Why LLMs can generate profound-sounding insights that users forget within 24 hours
• The Goldilocks zone of evidence: how much is enough, and when does more actually hurt you?
• Role confusion as a safety crisis: why being a companion and a therapist is clinically dangerous
  
  🎙️ Host: Paul Wicks | ProofStack Health
  🏢 Guest: Dr. Ali Darcy, Founder, Woebot Health
  🔗 Learn more: www.proofstack.health
  📬 Subscribe to Proof Points, A biweekly newsletter on evidence that makes digital health work proofpoints.proofstack.health

Digital health is littered with extraordinary claims that never collide with evidence. ADA Health is one of the rare exceptions.

In this episode of Prove It, host Paul Wicks sits down with Daniel Nathrath, founder and CEO of ADA Health, to unpack what it actually takes to build medical-grade AI over 15 years without cutting corners on safety, rigour, or evidence.

Daniel opens up about why he, a lawyer by training with a background in consumer internet, ended up building one of the world's most clinically validated symptom assessment platforms.

From the early decision to build for doctors first, to the hard lesson that healthcare is nothing like e-commerce, to the moment his Berlin office building became global news when a giant aquarium exploded in the lobby, this is a candid conversation about what building carefully actually looks like when the stakes are patient safety.

They cover:

• Why ADA chose to build a curated medical knowledge base over a pure LLM, and why that's now its biggest competitive advantage
  
• How ADA eliminates hallucinations entirely, and why pharmaceutical partners won't work with companies that can't say the same
  
• The "patient finder" model: shortening the diagnostic odyssey for rare disease patients from 7–10 years to something far more humane
  
• What 40 million assessments, a Class IIa medical device certification, and 20+ peer-reviewed papers actually mean in practice
  
• The real conversations Daniel plans to have at JP Morgan Healthcare, and why early patient identification is the future of pharma
  

If you care about what medical-grade AI trust at scale actually looks like, this is the episode to study.

Medical publishing looks objective from the outside, but behind the curtain it’s a high-pressure, incentive-driven system where great science can get buried and mediocre work can slip through.

In this special Prove It episode, Paul Wicks interviews a senior journal editor who remains anonymous (“The Secret Editor”) and explains how papers really move through the sausage factory: desk decisions, editorial boards, associate editors, peer review overload, and the surprisingly human judgments that shape what gets published.

They unpack the profit engine that reshaped scientific publishing, why “read the journal guidelines” is the #1 rule for getting accepted, and how the explosion in submissions has turned peer review into an arms race, ripe for gaming and burnout.

Finally, they tackle AI’s impact on submissions and publishing workflows, and why the deeper fix may be changing academic incentives, not just blaming publishers.



00:00 Meet the “Secret Editor”

02:14 The Dirty Secret: Scientists Keep Handing Publishers Power

04:37 How One Deal (Pergamon Press) Helped Create Today’s Publishing Machine

06:39 From Conferences to Peer Review: Why Journals Became the Battleground

08:57 #1 Rule to Get Published: Read the F!*%$ing Guidelines

12:51 The Submission Game: Acceptance Rates, Timelines, and “Cascade” Journals

17:33 Desk Rejections, Human Bias, and the 60-Manuscript Speed-Run

21:10 Open vs Blind vs Double-Blind Peer Review (And Why None Is Perfect)

24:44 How Bad Actors Game Peer Review, and Why Paying Reviewers Gets Messy

31:31 “Where Is My Paper?” Inside the Nightmare Workflow (20 Queues, 250 Manuscripts)

36:12 AI in Publishing: Some Editors Embrace It… Others Reject Anything “Near AI”

43:49 Blow It Up or Fix Academia? The Real Incentives Behind the Chaos

Digital health is drowning in bold claims, so what actually counts as evidence?

In this episode of Prove It!, Paul Wicks sits down with Dr. Acacia Parks (former Chief Science Officer at Happify, now VP of Regulatory Compliance & Engineering at Avania) to unpack the toughest questions in digital therapeutics: sham comparators, placebo effects, missing data, and what the FDA really expects.

They explore why dropout is “normal” in app-based trials, why sham design is both essential and fiercely guarded, and how the evidence needed for clearance differs from the evidence needed for adoption and reimbursement.

Then they tackle the elephant in the room: LLM “therapy” exploding at scale without regulation.

Finally, Acacia breaks down recent US shifts around wearables, the CMS TEMPO program and makes her case for the evidence standard she’ll defend every time: if you can’t beat placebo, what are we doing?