Sword

For most of history, access to high-quality healthcare has been a privilege of circumstance. Changing that takes bravery. That's why we partnered with Monocle to publish 𝘛𝘩𝘦 𝘍𝘶𝘵𝘶𝘳𝘦 𝘰𝘧 𝘏𝘦𝘢𝘭𝘵𝘩𝘤𝘢𝘳𝘦. Inside this limited-edition hardcover, world leaders, industry pioneers, and visionary clinicians share their boldest ideas for the road ahead, including essays and interviews from Patrick Conway, the CEO of Optum, Penny Dash, the Chair of NHS England, Marjorie Michel, Canada’s Health Minister, Vinod Khosla from Khosla Ventures, Chris Bischoff from General Catalyst, alongside our very own CEO Virgílio (“V”) Bento, and many more. Don’t just watch this movement from the sidelines, lead it. Request your copy now: https://lnkd.in/eMYf3HEr

A great idea is just the beginning. The future only arrives when someone is brave enough to back it. The Future of Healthcare, a limited-edition book by Sword and Monocle. Out tomorrow.

We are thrilled to welcome Caroline McGoldrick to Sword as our new Chief Strategy Officer. Caroline brings 25+ years of leadership from Aon and Mercer, where she scaled national specialty practice portfolios across pharmacy, clinical care, and leave solutions while advising C-suite leaders through complex health transformations. “Organizations don’t need another point solution, they need a fundamentally new model of care,” says Caroline. “It was clear that Sword is the undisputed industry leader delivering on that promise.” As companies demand unified solutions that lower costs and improve clinical outcomes, Caroline’s leadership will be instrumental in scaling our AI Care platform worldwide.

Earlier this week, we sat down with Abinue Fortingo of Brown & Brown to talk about choosing a metabolic health partner that actually gets members healthier. Abinue leads Brown & Brown's large-market population health team, so he knows exactly what to ask when a vendor makes its pitch. We pulled his sharpest advice from the conversation into the cheat codes below, from what "physician-led" should really mean for your members to how the best partners put their fees on the line. Swipe through, then watch the full recording on demand for Abinue's complete framework: https://lnkd.in/g6-aaeT8

This is healthcare’s Apollo moment. And we’re inviting everyone with a stake in healthcare to join us. This is the first time we’ve laid out our vision in full detail. In it, we cover the safety guardrails and principles that guide responsible AI in healthcare, along with the stakes and the work that lies ahead. Give it a read and tell us what you hope for the future of healthcare.

Big news for Canadians managing pain on disability leave. Recovery just got a lot more accessible. Manulife is partnering with Sword to bring AI Care physiotherapy straight to Group Benefits members on disability leave, helping members access physiotherapy more quickly and conveniently, and supporting faster recovery and return-to-work outcomes. No more choosing between waiting weeks for an appointment or driving across town for care. Members now get a licensed Canadian physiotherapist, a Health Canada-approved motion tracking tablet, and real-time progress tracking, all from home, and this is just phase one. In 2027, the program expands to every Group Benefits member with Extended Health Care coverage, opening access even wider across Canada. Read more in our newsroom: https://lnkd.in/ghPmT4Jv

Sword is partnering with Brazil’s Unified Health System (SUS), the world’s largest universal public healthcare system, serving more than 210 million people. The partnership was announced and signed with Brazil’s Ministry of Health during the 21st Brazilian Congress of Health Informatics, marking the start of a collaboration to explore how AI can help transform healthcare operations across SUS, from clinical triage and waiting list prioritisation to scheduling, clinical decision support, and prevention. This marks another major step in Sword’s work with national health systems. Over the last year, we’ve announced partnerships with the UK, Greece and Portugal, expanded our work with healthcare systems in Germany, and now we’re bringing that experience to Brazil. Given the scale of SUS, this will be Sword’s broadest collaboration with a national health system to date. This partnership gives us the opportunity to extend the impact of our AI technology to a healthcare system serving more than 210 million people. Read more in our newsroom: https://lnkd.in/ggzN6bCm

Today at 12:00PM ET, we're getting into how to think through your GLP-1 and cardiometabolic vendors. Abinue Fortingo of Brown & Brown walks through the framework he uses with employers, joined by Sword's Kevin Wang and Amanda Conway, MS, RDN Conway. Register here: https://lnkd.in/e2Qw6veY Can't make it? Register anyway and we'll send the recording.

Every era has a problem big enough to demand the very best of human ingenuity. In 1962, it was the original moonshot. Today, it's healthcare.

AI has made a fundamentally new model of healthcare possible. Delivering it is going to take the kind of speed and coordination we haven't seen since the space race.

Today, we’re laying out our vision in full, alongside the safety considerations and coordination it will take to get there. This is not something we can build alone. It will require the best of us, together. Join us. Link to the full manifesto in the comments.

New research from our AI Research team, accepted at COLM 2026: when an LLM judges its own outputs, even a fully objective rubric doesn't stop it from grading itself easier. We tested whether rubric-based evaluation, the method now used in benchmarks like HealthBench and a growing share of RL reward signals, is immune to self-preference bias. It isn't. In the worst case, a judge was 20x more likely to mark its own failed unit tests as passed than another model's. The bias held up even after controlling for how hard the outputs were to judge. Ensembling judges helped but didn't eliminate it. And on HealthBench, self-preference moved scores enough to reorder a leaderboard. Not every model family showed this bias equally, which suggests it isn't a fundamental limit of LLM judging but a trainable property. The paper digs into where the bias concentrates, why panels of judges only get you so far, and what actually helps. Read the full post by José Maria Pombal and Ricardo Rei here: https://lnkd.in/diqjmBX7