Healthcare Is Drowning in $1 Trillion of Paperwork. A Computer Use AI Agent Can Fix It. Will Anyone Let It?
American healthcare spends roughly $1 trillion every single year on administrative tasks. Not care. Not research. Not equipment. Administration. Billing. Prior authorizations. Faxes. Literal faxes. McKinsey estimates that $265 billion of that is pure, recoverable waste. And yet in 2026, the average physician still spends more than 28% of their working hours doing EHR data entry and paperwork instead of seeing patients. That's not a staffing problem. That's not a budget problem. That's a 'we refuse to automate the obvious' problem. The technology to fix this exists right now. A proper computer use AI agent can navigate any EHR system, fill out any form, handle prior authorizations, and process insurance data without a single human clicking a single button. The real question isn't whether AI can do this. It's why healthcare organizations keep choosing not to let it.
The Prior Authorization Hellscape Nobody Talks About Enough
Here's a number that should make every healthcare administrator physically ill. The average physician's practice handles nearly 40 prior authorization requests per week. Their staff spends 13 hours completing them. Thirteen hours. That's basically a part-time employee whose entire job is begging insurance companies for permission to do medicine. According to the AMA's own survey data, patients are so exhausted by the prior auth process that many of them just give up and abandon care entirely. They don't get the treatment. Not because it was denied, but because nobody had the bandwidth to fight for it. And what's the insurance industry's answer to this crisis? More AI. Specifically, AI that denies claims faster. UnitedHealthcare faced a class-action lawsuit alleging its AI algorithm wrongfully denied claims at massive scale. The AMA reported in early 2025 that physicians are genuinely afraid that unregulated insurance AI is making prior authorization denials worse, not better. So we have a situation where the payer side is deploying computer-using AI aggressively to say no faster, while the provider side is still doing everything by hand. That asymmetry is costing patients their health.
What Your Doctors Are Actually Doing All Day (It's Not Medicine)
- ●28.4% of physician time goes to EHR input and administrative work, per peer-reviewed ecological momentary assessment data published in PMC
- ●Physicians are checking their EHR systems during paid time off, including vacations, according to a March 2025 JAMA Network Open study
- ●Healthcare administrative spending in the US hits approximately $1 trillion annually, with $265 billion identified as directly saveable through known interventions
- ●A 2025 Nature study analyzed burnout data from 43,026 healthcare workers and found administrative burden is a primary driver across every department
- ●Doctors spend 13+ hours per week on prior authorizations alone, time that could serve roughly 5 to 7 additional patients per physician per week
- ●The EHR still follows doctors home, with significant after-hours documentation time reported in August 2025 AMA data, destroying work-life balance across the profession
Of the roughly $1 trillion in annual healthcare administrative spending, the US could save $265 billion every year with automation that already exists. That's not a prediction. That's a McKinsey analysis from 2021 that healthcare has spent five years mostly ignoring.
Why RPA Bots Are Not the Answer (And Never Were)
The healthcare industry did try to automate. In the 2018-2022 window, everyone bought RPA. Robotic Process Automation tools from UiPath, Automation Anywhere, and Blue Prism flooded hospital IT departments with promises of efficiency. And to be fair, some narrow tasks got faster. But RPA has a fundamental, structural problem that nobody in the sales deck mentioned. RPA bots are brittle. They follow rigid, pre-programmed scripts. The moment an EHR vendor updates their UI, or a form adds a new field, or a workflow changes slightly, the bot breaks. And in healthcare, things change constantly. Regulations change. Payer portals change. Epic releases updates. Cerner does something weird. The result is that healthcare IT teams spend enormous time maintaining RPA bots that were supposed to save them time. It's automation theater. You get the press release, the vendor case study, and then a 2am call when the billing bot crashes and nobody knows why. Real computer use AI is categorically different. A computer use agent doesn't follow a script. It sees the screen the way a human does, reasons about what it's looking at, and figures out what to do next, even if the interface changed since yesterday. That's not a subtle improvement. That's a completely different category of tool.
The Ethical Minefield Everyone Is Tiptoeing Around
Let's be honest about something the enthusiast press keeps glossing over. AI automation in healthcare has a real ethics problem, and pretending otherwise is naive. The UnitedHealthcare AI controversy isn't a bug in the system. It's what happens when you deploy automation to optimize for cost reduction without building in accountability. A Stanford HAI study from June 2025 found that AI mental health chatbots may not only be less effective than human therapists but could actively contribute to harmful stigma. These are not hypothetical risks. They're documented failures happening right now. So when I say healthcare needs aggressive computer use automation, I mean the administrative layer, the paperwork, the data entry, the form routing, the scheduling, the billing reconciliation. The stuff that doesn't require clinical judgment. Not the clinical decision-making. Not the diagnosis. Not the therapy session. The distinction matters enormously, and organizations that blur it are going to cause real harm and set back legitimate automation by a decade. The goal is to free clinicians from administrative slavery so they can do more of the actual human work that AI genuinely cannot replace.
Why Coasty Exists and Why Healthcare Should Care
I've watched a lot of tools get pitched as the solution to healthcare's admin crisis. Most of them are either glorified chatbots with a healthcare logo slapped on, or RPA rebranded with 'AI' in the name. Coasty is neither. It's a genuine computer use AI agent, meaning it controls real desktops, real browsers, and real terminals the same way a trained human employee would. It scored 82% on OSWorld, which is the gold-standard benchmark for evaluating how well an AI agent can actually operate a computer in the real world. For context, Claude's computer use scored 61.4% on OSWorld. Coasty is not close to the competition. It laps it. For healthcare specifically, that performance gap matters more than in almost any other industry. Your EHR is messy. Your payer portals are inconsistent. Your workflows have exceptions and edge cases that a scripted RPA bot will choke on. A computer use agent that actually understands what it's looking at, like Coasty does, handles those exceptions the way a smart new hire would. It figures it out. Coasty runs on a desktop app, spins up cloud VMs, and can run agent swarms for parallel task execution, meaning you can process a hundred prior authorizations simultaneously instead of one at a time. There's a free tier to start. BYOK is supported if your compliance team has opinions about API keys. If you're a healthcare org still paying people to copy-paste data between systems in 2026, the conversation you need to have with your CTO starts at coasty.ai.
Here's where I land on this. Healthcare's administrative crisis is not a mystery. The causes are documented. The costs are quantified. The solutions exist. What's missing is the organizational will to actually deploy them, and to deploy them in the right places. Stop using AI to deny claims faster and start using it to eliminate the paperwork that's burning out your staff and delaying patient care. Stop buying RPA bots that break every time Epic sneezes. Start using a real computer use AI agent that can handle the actual complexity of your environment. The $265 billion in recoverable waste isn't going anywhere on its own. Every year you wait, that's another 13 hours per week per physician spent on prior authorizations instead of patients. Another doctor checking their EHR on vacation. Another patient abandoning care because the prior auth process was too exhausting to fight. The tools are ready. The benchmark data is clear. The ROI math is embarrassingly obvious. The only thing left is the decision to act. Go to coasty.ai and see what a computer use agent actually looks like when it works.