Medicare is using AI to approve claims. The result has been ‘alarmingly high denial rates.’
THE SO WHAT
An AI claims pilot that drives “alarmingly high” denial rates and vendor warnings about unreadiness is a case study in what happens when you automate adjudication faster than you harden policy and appeals. Any operator deploying AI into rights- or benefits-determining workflows should slow down, add human-in-the-loop checkpoints, and assume legal discovery will surface every design decision you make.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIThe big AI 'slowdown' battle explained — 5 things you need to know about Trump's AI regulation battle with OpenAI's Sam Altman, Anthropic's Dario Amodei, and even Elon Musk
When frontier lab CEOs and the White House are publicly split on existential AI risk, AI governance becomes a political wedge, not just a technical debate. If you’re building on US-regulated AI, assume rules may oscillate with administrations and bake regulatory volatility into product and infra planning.
Applied AIIntroducing Gemini 3.8 Live and 3.8 Live Extended Thinking
Gemini 3.8 Live and its Extended Thinking mode push toward persistent, conversational agents that can reason over longer horizons. For operators, the question shifts from “can the model answer this?” to “what workflows can we safely hand over end-to-end, and how do we supervise them over time?”
Applied AIMeta expands subscription push with new AI-focused plans
Meta One bundling AI access with premium features across Facebook, Instagram, and WhatsApp makes AI a subscription upsell lever, not just a retention tool. If you run a consumer or prosumer product, revisit your pricing grid — AI features are now a line item customers expect to see and evaluate explicitly.
Anthropic says it created an AI 'starter pack' for small businesses
Anthropic’s Claude ‘starter pack’ for small businesses — plus workshops to drive adoption — shows labs going direct to the SMB layer instead of waiting for SaaS resellers. If you sell into SMBs, expect customers to arrive with pre-baked AI expectations and some backlash risk; your value will be workflow depth and trust, not raw model access.