Your Doctors Are Doing 15 Hours of Paperwork a Week. A Computer Use Agent Could Fix That Today.
Fifteen hours. Every single week. That's how much time the average physician spends on EHR work outside of normal clinic hours, according to Healthcare IT News. They have a name for it: 'pajama time.' Doctors, who spent a decade in school to save lives, are sitting in bed at 11pm clicking through insurance portals and copy-pasting billing codes. Meanwhile, the healthcare industry is sitting on a verified $20 billion annual savings opportunity from automation, and most of it is going nowhere. We are in 2026. The tools exist. The benchmarks are real. The ROI is not even debatable. So why is your hospital still running on manual workflows that would embarrass a 2012 startup?
The $20 Billion Problem Nobody Wants to Solve
The CAQH Index dropped a report in early 2025 that should have been front-page news everywhere. The U.S. healthcare industry has a $20 billion opportunity sitting right there, in administrative waste, just waiting to be cut. Not through some futuristic moonshot. Through automating the repetitive, soul-crushing transactions that staff do manually every single day: eligibility checks, prior authorizations, claim submissions, referral tracking. These are not complex medical decisions. They are data-entry tasks dressed up in clinical clothing. And yet hospitals keep hiring more administrative staff to do them, while physician burnout hits record levels and patient wait times stretch into weeks. A 2025 study in JAMA Network Open confirmed that ambient AI scribes alone reduce documented burnout and after-hours charting. That's just one narrow slice of what full computer use automation can actually do. The irony is brutal: healthcare is the industry that needs automation most, and it's one of the slowest to actually deploy it at scale.
Prior Authorization Is a War Zone, and AI Is Making It Worse in the Wrong Hands
Here's where it gets genuinely infuriating. The AMA published a press release in February 2025 warning that AI is actually increasing prior authorization denials. Health insurers are deploying AI to reject claims faster, not to help patients get care faster. CMS launched the WISeR model in mid-2025 to test AI-driven prior authorizations across Medicare, and doctors in six states immediately raised alarms about more denials, more delays, and zero transparency into why an algorithm said no. Stateline reported in December 2025 that thousands of Medicare patients could have claims denied by AI with no clear human review. This is what happens when the wrong side of the industry gets to the technology first. Insurers use computer-using AI to automate denial. Providers are still faxing documents back. The asymmetry is obscene. The fix is not to slow down AI adoption in healthcare. The fix is for providers and health systems to arm themselves with the same class of tools and fight back, automating their own submissions, appeals, documentation, and compliance workflows so they're not bringing a fax machine to a gunfight.
Physicians report 15 hours of 'pajama time' per week on EHR and admin tasks. That's 780 hours a year, per doctor, spent not treating patients. At average physician compensation rates, that's roughly $140,000 in productive time vaporized annually, per person, on paperwork.
Why RPA Alone Was Never the Answer (And Why Healthcare Kept Buying It Anyway)
Ask any health system IT director about their UiPath or Blue Prism deployment from 2019 to 2023. Then watch their face. Traditional RPA (robotic process automation) promised to automate healthcare workflows, and it delivered, sort of, until the interface changed, the portal updated, or the workflow had one unexpected variation. RPA bots are brittle. They follow scripts. They break when the real world doesn't cooperate, which in healthcare is basically always. A prior auth portal update, a new field in the EHR, a slightly different PDF layout from an insurer, and suddenly your 'automated' workflow needs a developer to fix it. That's not automation. That's a fragile robot with a very expensive maintenance contract. The CAQH Index noted that even with years of RPA investment, manual workflows still dominate most administrative healthcare transactions. The industry spent billions on automation that couldn't actually adapt. What healthcare actually needs is a computer use agent that sees the screen the way a human does, reasons about what it's looking at, and figures out what to do next, even when the interface changes. That's a fundamentally different category of tool.
What Real AI Computer Use Looks Like in a Health System
Forget the chatbot demos. Forget the 'AI assistant' that summarizes notes and calls it a day. A real computer use agent operates the actual software your team already uses. It opens the EHR. It reads the patient record. It navigates to the insurer portal, fills out the prior auth form, attaches the clinical documentation, and submits it. Then it checks the status the next morning and flags anything that needs human review. No API integration required. No six-month implementation project. No dedicated developer to maintain the bot when the portal changes. This is not theoretical. The technology exists right now, and the benchmarks prove it. OSWorld, the gold standard for measuring how well AI agents can actually operate real computer software, has become the proving ground for this category. Systems that score well on OSWorld can handle the messy, multi-step, real-world workflows that healthcare actually runs on. Scheduling across systems, insurance verification, billing code lookups, discharge documentation, referral coordination. All of it. The question is not whether AI computer use can do this. The question is why your health system is still waiting.
Why Coasty Is the Tool Healthcare Operators Should Actually Be Looking At
I've looked at the benchmarks carefully, and Coasty is the computer use agent I'd put in front of a health system operations team without hesitation. It scores 82% on OSWorld, which is the highest of any computer use agent available right now, and that gap matters in healthcare where workflows are complicated, legacy software is everywhere, and errors have real consequences. Coasty controls actual desktops, real browsers, and terminals. It doesn't need your EHR vendor to build an API. It doesn't require a six-figure implementation contract. It works the way a trained human works, by looking at the screen and figuring it out, except it doesn't get tired at hour twelve of eligibility verification. The agent swarm capability means you can run parallel workflows simultaneously, checking insurance eligibility for a hundred patients while simultaneously following up on outstanding claims and pre-filling prior auth requests. There's a free tier if you want to actually test it before committing, and BYOK support if your compliance team has opinions about where the model runs. For healthcare operations teams that are genuinely serious about cutting administrative overhead in 2026, this is the benchmark leader. Not by a little. By a lot.
Here's my honest take: healthcare administration is a solved problem that the industry refuses to solve. The stats are not ambiguous. Fifteen hours of pajama time per physician per week. Twenty billion dollars in automatable waste. Billing error correction eating 20% of some workers' entire month. And the response from most health systems is another committee, another vendor evaluation, another pilot that never scales. That's a choice. It's a choice that burns out doctors, delays patient care, and hemorrhages money that could go toward actual medicine. The tools that can fix this are not experimental. They're benchmarked, deployed, and available today. A computer use agent that actually scores at the top of OSWorld, like Coasty, can start touching real workflows this week, not after your next fiscal year planning cycle. Go to coasty.ai, run the free tier on one workflow, and see what happens. If you're still manually processing prior auths after that, I genuinely don't know what to tell you.