Blogs

Most telehealth platforms in production today were built between 2020 and 2022. They were built fast under clinical pressure, with minimal EHR integration, compliance scoped for basic HIPAA, and no architecture for the AI capabilities that were three years away. At the time, that was the right call. Patients needed access. Providers needed tools. Speed was the only metric that mattered.

The downstream revenue loss from referral leakage reaches $800,000 to $900,000 per physician per year. Not from billing errors. Not from denied claims. From referrals that were written and never completed. At a 100-provider multi-specialty group, that is $78 to $97 million in annual revenue that left the network silently, without a single denied claim to flag it.

Physicians complete an average of 39 prior authorization requests per week and spend 13 hours on the process (AMA Prior Authorization Survey). At a 500-bed health system, that translates to more than $250,000 in annual administrative overhead from manual processing alone based on typical PA volumes, staffing requirements (3–5 FTEs dedicated to manual PA processing), and per-transaction administrative costs before denied claims, before rework, before the staff turnover that follows sustained administrative overload.

Claude Code is configurable. The model it uses for reasoning and code generation sits underneath the interface and you can swap it. Most teams using Claude Code have never tested what happens when you do.

Between October 2025 and March 2026, Anthropic, OpenAI, and Amazon Web Services each launched a purpose-built healthcare AI platform. Not general-purpose AI with a compliance disclaimer. Purpose-built, HIPAA-aligned products targeting clinical operations, revenue cycle, life sciences, and patient access.

Ask any director of institutional research or head of analytics at a mid-size university how their backlog looks. The answer is almost always some version of the same thing: too many requests, not enough people, and a queue that has been growing longer for at least two years.

You already know the salary number. You've seen the Glassdoor ranges. You know that a strong senior data engineer in the US runs $135K–$170K base depending on market, stack, and experience. If you're in a high-cost metro or competing with FAANG for talent, you've seen higher.

It's Monday morning. Your lead data engineer opens Slack and there are already 11 messages waiting. A Prefect flow failed overnight. A dbt model broke because someone upstream changed a schema without telling anyone. There's a ticket from Finance asking why last week's revenue numbers don't match what the dashboard showed on Friday. And somewhere in the backlog, buried under 23 maintenance items is the ML initiative you announced to the board last quarter. 

Your competitor shipped an AI assistant last quarter. Your board has asked three times in the last two meetings when you're shipping yours. Your VP of Product has a roadmap that includes "AI integration" in the next sprint cycle.