Healthcare Is Drowning in Paperwork While Your AI Computer Use Agent Sits Idle

The $35 Billion Problem Nobody Wants to Say Out Loud

Let's talk about prior authorization, the single most maddening administrative ritual in modern medicine. Health Affairs put the total annual cost of prior authorization at $35 billion across the US healthcare system. Physicians alone absorb $26.7 billion of that. Every week, doctors and their staff are filling out the same forms, calling the same insurance lines, uploading the same documents into portals that look like they were built during the Obama administration. And here's the part that should make you furious: nearly 20% of healthcare workers spend more than 20 hours a month just correcting billing errors, according to a 2025 Healthcare IT News survey. That's half a work week, every single month, fixing mistakes that a computer should never have made in the first place. The American Hospital Association reported hospitals spent more than $1 trillion paying healthcare workers in 2025. A significant slice of that trillion dollars is funding humans to do tasks that are, frankly, beneath them. Tasks that a well-built computer use agent could handle at 2am without complaining.

Why Your Current Automation Is a Fraud

• ●Traditional RPA bots are brittle. They break the moment a payer portal updates its UI, which happens constantly. Then someone has to fix the bot. Then it breaks again.
• ●More than 95% of businesses that implement RPA fail to scale it beyond a handful of processes, per industry analysis from late 2025.
• ●RPA can't handle judgment calls. If a form has an unexpected field or a popup appears, the bot stops dead. A human gets paged. The whole point collapses.
• ●Legacy automation requires IT teams to hard-code every workflow step. In healthcare, workflows change constantly due to payer rule updates, regulatory shifts, and software upgrades.
• ●The AMA found in February 2025 that physicians are now worried insurers are using AI to increase prior authorization denials automatically, while providers are still fighting back with manual processes and outdated bots. That's an asymmetric war nobody on the provider side can win.
• ●Menlo Ventures' 2025 State of AI in Healthcare report found that 37% of healthcare organizations say clinical documentation is their top automation priority, yet most are still relying on approaches that require constant maintenance and IT babysitting.

The Burnout Crisis Is an Automation Crisis in Disguise

Physician burnout is treated like a mental health problem. It's not. It's an operational problem. The AMA's 2025 data shows that EHR work follows doctors home, with nearly an hour of EHR time logged on weekend days alone. A JAMA Network Open study from October 2025 confirmed that physician groups with lower administrative burden have measurably better retention, higher quality care scores, and lower burnout rates. We know the cause. We know the solution. The gap is execution. The old argument was that automation couldn't handle the complexity of healthcare workflows, the edge cases, the multi-step portal navigation, the documents that need to be read and interpreted before being filed somewhere else. That argument was valid in 2022. It's not valid anymore. What changed? Computer use agents that can actually see a screen, read context, make decisions, and navigate any interface without needing a hard-coded script. This is categorically different from RPA. An AI computer use agent doesn't need to know the exact pixel location of a button. It understands what the button does.

What Real Computer Use AI Actually Does in Healthcare

Here's what a modern computer use agent can handle that traditional automation never could. It can log into a payer portal, read the current prior authorization requirements, fill out the form based on patient record data, upload supporting documents, and confirm submission. All of it. Without a template. Without breaking when the portal gets a redesign. It can cross-reference an EHR entry against a billing code database, flag discrepancies, and draft a corrected claim. It can monitor multiple insurance portals simultaneously for status updates and route alerts to the right staff member. It can handle the long tail of tasks that make up the administrative nightmare, the stuff that's too variable for old-school RPA but too repetitive to justify a human doing it. This is what computer-using AI was built for. Not chatbots. Not API integrations that only work when systems cooperate. Actual desktop and browser control, the way a human would do it, but faster and without the burnout.

Why Coasty Is the Computer Use Agent Healthcare Actually Needs

I've looked at the options. Anthropic's computer use is interesting research. OpenAI's Operator is fine for simple consumer tasks. UiPath is still pitching RPA with an AI coat of paint. None of them are purpose-built to handle the kind of complex, multi-step, judgment-heavy workflows that healthcare admin demands. Coasty is different. It scored 82% on OSWorld, the hardest benchmark for computer use agents, and nothing else is close. That score matters in healthcare because OSWorld tests exactly the kind of real-world, unpredictable task navigation that prior auth portals and EHR systems throw at you every day. Coasty controls real desktops, real browsers, and real terminals. Not API wrappers. Not simulated environments. It runs agent swarms for parallel execution, so you're not waiting in a queue while one bot slowly works through a stack of authorizations. It has a free tier and supports BYOK, so you're not locked into a pricing model designed to extract maximum revenue from health systems that are already underwater. If you're a hospital administrator, a practice manager, or a health tech team that's tired of explaining to physicians why the bot broke again, coasty.ai is where I'd start.