Day 3 swapped hype for hard edges: preventative “health OS” models, media’s pivot from pageviews to queries, explainable finance AI, instant payments across borders, and a reminder that human judgment, not speed, differentiates in the AI era.
Health, productized: toward a personal “OS”
Prenuvo’s CEO Andrew Lacy and investor Tony Fadell framed preventive care as a product experience people want to repeat, annual whole-body MRI plus AI insights, delivered like a premium consumer service. Fadell recounted a friend’s stage-one brain cancer caught by a trial scan, one of “six friends whose lives have been saved”, he said, before describing how longitudinal imaging becomes behavior-changing evidence: “A big list isn’t necessarily a strong list… what’s most powerful is seeing change over time.”
Lacy was blunt about psychology: people avoid diagnostics because every story they hear is bad; the company is trying to “change the definition of disease” by creating good stories of early detection and reversal. One striking finding from Prenuvo’s spine model: the average 25-year-old spine looks like a 40-year-old’s, likely from hours looking down at phones, actionable now, decades before symptoms.
The service design matters as much as the scan: a calm, non-clinical visit, 40–45 minutes, media or music during the MRI, and a mobile app with images and reports. It’s prevention delivered like modern software: personal, repeatable, and sticky.
Media after the homepage: queries are the new unit of value
“In 2025 the web is no longer built for attention, it’s built for answers”, argued Brooke Hartley Moy (Infactory). With AI interfaces mediating information, publishers must stop thinking like media companies and start operating like data companies: structure archives, annotate, and license for machine precision. Monetization shifts from pageviews to micro-transactions on queries, including those generated by agents, not just by humans. The archives (AP-scale text, image, video) become high-value, rights-aware corpora for fine-tuning and retrieval, especially for niche, local, real-time content that gen-AI cannot cheaply invent.
Blocking bots is a short-term reflex, Moy said; the strategic move is industry coalitions, rights standards, and schema that make content machine-readable with provenance. Inside newsrooms this means staffing up data engineering and treating the CTO as revenue-critical.
Finance AI, unboxed: accuracy, explainability, and fraud at speed
Tamir Hazan (Sunbit) opened the black box: approving ~90% of applicants profitably demands architectures beyond tree models, think visual-style neural nets that treat credit files as structured “images”, plus generative counter-fraud (simulate attacks you haven’t seen) and reinforcement learning for service flows with a human-in-the-loop. The catch: finance data is sparse, shifts with macro shocks (e.g., stimulus on/off), and must be explainable to developers, regulators, and consumers. Continuous training, reward design for agents, and first-party labeling are the hidden costs of “AI at work”.
The zero-delay economy: instant, 24/7, and safer
Payments leaders from ClearBank and Brite were clear: real-time is table stakes, 97% of ClearBank’s UK payments clear in ~2 seconds, but the hard work is reconciliation and fraud that must also run in real time. The EU bars slowing instant payments for fraud checks, so risk moves earlier in the stack (account opening, biometrics, behavioral analytics). Europe’s fragmentation persists (SEPA vs. domestic rails, seven non-euro markets), yet Pay by Bank is rising because it’s instant, lower-cost, and strongly authenticated via the user’s own banking app.
On cross-border, panelists pointed to stablecoins/tokenized deposits as the infrastructure answer that bypasses incompatible rails, settlement that’s faster, cheaper, and traceable. ClearBank’s new USDC/EURC work with Circle hints at banks embracing blockchain under regulated wrappers.
Inclusion, if we design for it: stablecoins in the Global South
There are 1.4B unbanked people. For remittances and savings in volatile currencies, stablecoins are jumping the gate: near-instant transfers, predictable value in digital dollars, and new access to digital services otherwise blocked by local rails. Speakers from Zepz and Transfero stressed the real world: legacy cross-border can take a week and land unpredictably; stablecoins compress time, fees, and uncertainty, if on/off-ramps, KYC/AML, and tax clarity keep pace. The regulatory arc (US, EU MiCA, UK) is bending toward formalization, which should bring mainstream scale.
Humans differentiate: the creative edge in an AI-saturated workflow
Closing the day, Jean Philippe Rosier (Perestroika) offered a pragmatic creative system: let AI produce at speed, but judge with four lenses, relevance, feasibility, non-obviousness, resonance. Treat models as co-intelligence, not crutches: set intention, frame with constraints, apply human judgment, iterate. “Acceleration without discernment is just noise”, he said; the future belongs to teams that think best with AI, not those who prompt fastest.
The signal from Day 3
- Preventive health is becoming a consumer product with AI as the explainer and motivator. Evidence beats advice.
- Publishing is moving from ad-supported attention to rights-aware query commerce, data engineering is the new newsroom.
- Finance AI must be auditable and robust to regime shifts; fraud defense is becoming generative and game-theoretic.
- Zero-delay money needs instant rails and instant trust; stablecoins look like the pragmatic bridge for cross-border.
- Creativity scales when humans decide what not to ship. AI accelerates; operators differentiate.
Why it matters
Day 3’s signal is executional, not speculative: healthcare becomes a repeatable product with AI as the explainer; media stops chasing clicks and monetizes answers; finance AI proves value only when it’s auditable; instant payments matter only if fraud controls are instant too; and stablecoins turn cross-border pain into programmable money, especially for the unbanked. The competitive edge now is judgment at speed: teams that can prove provenance, log decisions, and pause (via “I don’t know”) will win trust; teams that ship prototypes, measure in production, and iterate on real-world feedback will win markets. In short, the next cycle rewards operators who turn trust into infrastructure, latency into product, and AI into outcomes.
Catch up with our complete Web Summit 2025 series:
Web Summit 2025 Opens with a Power Shift: Pix, Poland and Sovereign AI Take Center Stage
Web Summit 2025 Day 1: On the Ground in Lisbon – AI Grows Up and Gets Practical
Web Summit 2025 Day 2: AI Steps Off the Screen and Into the Real World
Your writing is a true testament to your expertise and dedication to your craft. I’m continually impressed by the depth of your knowledge and the clarity of your explanations. Keep up the phenomenal work!