UiPath Had a 30% Stock Crash and a Securities Lawsuit. Maybe AI Computer Use Agents Were Right All Along.
Manual data entry costs U.S. companies $28,500 per employee every single year. That stat is from a 2025 report and it should make every operations leader feel sick. The promise of RPA, and UiPath specifically, was that we'd solved this problem. We hadn't. We'd just moved the pain somewhere else. Somewhere more expensive, more brittle, and a lot more embarrassing. UiPath's stock dropped 30% in a single day in May 2024 after a disastrous earnings report, its CEO walked out the door, and the company ended up facing a securities fraud lawsuit. Meanwhile, AI computer use agents are quietly doing everything RPA promised, without the fragility, without the six-figure implementation fees, and without the maintenance nightmare that eats your IT team alive. This isn't a hot take. This is just what happened.
The RPA Dream Was Always Kind of a Lie
Here's how RPA was sold to you: deploy a bot, it clicks through your apps like a human would, you save thousands of hours, everyone goes home happy. Here's what actually happened: your bot worked great for three weeks, then your ERP vendor pushed a UI update, and the bot started clicking the wrong button and corrupting records at 3am while nobody was watching. This is not a hypothetical. It's the defining experience of enterprise RPA adoption. Ernst and Young found that RPA projects fail at a rate of roughly 50%. Forrester found that 60% of RPA teams spend most of their time on maintenance, not new automation. Not building. Not expanding. Just keeping the lights on for bots that break every time an app sneezes. The core problem with UiPath and traditional RPA is architectural. These tools work by recording pixel coordinates and UI element selectors. They're essentially macros with better marketing. The moment the underlying application changes, the bot is blind. It doesn't understand what it's doing. It just knows where to click. That's not intelligence. That's a very expensive, very fragile script.
What the Numbers Actually Say About UiPath Right Now
- ●UiPath stock fell 30% in a single session after disappointing May 2024 earnings, wiping out billions in market cap overnight
- ●The company faced a securities fraud class action lawsuit in the Southern District of New York, with investors alleging they were misled about UiPath's business and revenue prospects
- ●UiPath's own annual report lists AI agents as an existential risk to its core business model, which is not something a healthy company writes
- ●30-50% of RPA projects fail before they ever deliver ROI, according to multiple industry analysts including Forrester and Gartner
- ●Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027, but the ones that survive will eat RPA's lunch entirely
- ●UiPath's revenue growth is now guided at 9% for fiscal 2026, which is not a number that justifies a $6 billion valuation in an AI-first world
- ●56% of employees report burnout from repetitive data tasks, meaning the human cost of not automating correctly is also enormous
"Maintenance costs approach development costs. Bots break when applications update. Business teams work around broken automation." That's not a critic talking. That's a direct description of the RPA lifecycle from an industry analysis. You pay to build it, then you pay again to keep it alive, then your team quietly stops using it and goes back to doing things manually. Congratulations, you spent six figures to go in a circle.
Why AI Computer Use Agents Are a Completely Different Category
A computer use agent doesn't memorize where the button is. It sees the screen the way a human does, reads the context, and figures out what to do next. Change the UI, move the button, rebrand the app entirely. The agent adapts. This is the fundamental architectural difference that makes AI computer use not just a better version of RPA, but a different thing altogether. OpenAI's computer-using agent scored 38.1% on OSWorld when it launched in January 2025. Anthropic's Claude has been creeping up with each model release. These are real benchmarks on real desktop tasks, not synthetic demos. The OSWorld benchmark tests agents on 369 real computer tasks across actual operating system environments. It's the closest thing the industry has to an honest test of whether a computer use agent can actually do your work. And the scores are climbing fast. The agents that are winning on OSWorld aren't winning because they have better macros. They're winning because they actually understand what they're looking at.
The UiPath CTO Said the Quiet Part Loud
In May 2025, UiPath's own CTO went on record saying AI agents will not take over the automation world. That's a fascinating thing to say when your company's entire pivot strategy is built around, wait for it, AI agents. UiPath is trying to bolt agentic capabilities onto a platform that was designed for a fundamentally different paradigm. It's like putting a Tesla drivetrain into a 2003 Camry and calling it an EV. The bones are wrong. The Reddit thread titled 'Not popular opinion: UiPath and similar tools are dead' from September 2025 has hundreds of upvotes and a comment section full of RPA practitioners who are tired of defending the status quo. These aren't AI hype bros. These are people who have spent years building UiPath automations and are now watching AI computer use agents do in an afternoon what took them weeks to script and test. The community knows. The market knows. The only people who don't seem to know are the ones who already bought the enterprise license.
Why Coasty Exists
I'm going to tell you about Coasty the same way I'd tell a friend at a bar: it's the best computer use agent available right now, and I can prove it. Coasty scores 82% on OSWorld. For context, OpenAI's computer-using agent launched at 38.1%. Anthropic's best scores are still climbing toward that number. 82% is not a rounding error advantage. It's a different class of performance. What Coasty actually does is control real desktops, real browsers, and real terminals. Not API wrappers pretending to be automation. Not bots that break when a dropdown changes. An actual computer-using AI that sees your screen and does the work. You get a desktop app, cloud VMs, and agent swarms for parallel execution, meaning you can run multiple tasks at once instead of waiting for a single bot to plod through a queue. There's a free tier if you want to test it without a procurement process. BYOK is supported if you're particular about your models. The reason Coasty exists is exactly the problem this post is about. RPA was always a workaround. Computer use agents are the actual solution. The gap between 'record a macro' and 'understand a screen' is the gap between UiPath and what Coasty does at coasty.ai.
Here's my honest take after looking at all of this: UiPath isn't going to zero tomorrow. Large enterprises have too much sunk cost, too many internal champions, and too much inertia to rip it out next quarter. But the trajectory is obvious. A 30% stock crash, a securities lawsuit, a CEO exit, and a CTO who goes on record trying to talk down the technology that's eating his company's market share. These are not the signals of a platform with a bright future. They're the signals of a category in decline. If you're evaluating automation today, and you're choosing between spending months scripting brittle bots that your team will spend half their time maintaining, versus deploying a computer use agent that actually understands what it's doing, the choice is not complicated. The $28,500 per employee you're losing to manual work every year doesn't care which tool you're loyal to. It just keeps adding up. Go try Coasty at coasty.ai. The free tier exists for exactly this moment.