Your AI doctor, powered by real doctors & leading medical evidence. Get free primary care and unprecedented health insights.
External link for Lotus AI
San Francisco, US
Lotus AI reposted this
For the past 12 years, Forbes and TrueBridge Capital Partners have worked together to highlight the next generation of unicorns. We're incredibly proud that Lotus AI and Outtake made this year's list. Congrats to KJ Dhaliwal, Alex Arjun Dhillon and their teams.
Lotus AI reposted this
Most medical AI evaluations ask whether a model can answer a benchmark question or generate a plausible differential. We wanted to study something harder and more operationally honest: when our AI proposes a real care action inside a live primary care workflow, does the physician agree? Today we're publishing our public research at Lotus AI. Across 1,678 AI-suggested care actions for 244 patients (follow-up messages, lab orders, and prescriptions generated by our physician copilot, Lotus Cortex), here's what we found: → Among actions that reached a final approve-or-reject decision, physicians approved 98%. → Action types behaved very differently. Follow-up messages were the most consistently accepted. Prescriptions carried the highest rejection rate at 10.2%, which is exactly where you want stricter guardrails. → Lab orders improved with iteration. More revision cycles between Cortex and the physician correlated with higher approval (Spearman ρ = 0.20, p < 0.001). The physician-agent loop is doing real work. → The hardest problem isn't rejection. It's pending. ~30% of proposed actions never reached a final verdict, which is the real operational bottleneck for clinical AI. A note on why we're sharing this: Our mission is to build the best clinical AI in the world, and safety has to be load-bearing, not marketing. That means evaluating our systems on what physicians will actually act on, being public about the methodology and its limits, and treating clinician agreement as a first-class signal alongside benchmark performance. To the engineers, researchers, and physicians thinking seriously about AI in healthcare: we welcome your reactions and critiques. This is the first of many. Read the full paper: https://lnkd.in/gwqEwrFa
Lotus AI reposted this
I polled over 1,000 people and 76% said their biggest frustration with healthcare in the US is high costs. It's expensive because the thing value of time is high and doctors' time is scarce. The work that moves the needle for complex patients is slow and detailed: 👉Pulling fragmented records together 👉Cross-referencing symptoms with the latest research 👉Identifying patterns that only emerge when you look at everything at once That doesn't happen in 15 minutes. So patients are left with no answers, or worse, wrong ones. This is the problem AI was built to solve. Not replacing doctors. Doing the time-intensive work clinicians can't afford to do at scale, then handing the results to a real doctor who can make the final call with a clear and concise summary of the full picture. I've recently been working with Lotus AI, and I've been shocked at how much of the tedious work can be automated, and still connected to the services patients need, instead of being another tool they have to use. Lotus is a platform providing free primary care powered by AI working with real physicians. It... 🪷Compiles all medical records into one place 🪷Identifies patterns across complex, fragmented health data 🪷Cross-references symptoms and medications with peer-reviewed research 🪷Created comprehensive summaries for real doctors to review 🪷Sends test orders and prescription requests to licensed clinicians There's a lot of discussion around what AI should and shouldn't be used for in healthcare. But if the biggest frustration is cost, and the biggest cost is time, then the question is really about what deserves a doctor's time and what's been wasting it.