UiPath Is Losing to AI Agents and Its Own Lawsuit Proves It

The UiPath Story Nobody Is Telling Out Loud

In September 2025, a securities fraud class action lawsuit was filed against UiPath. The complaint alleged that UiPath concealed the truth that customer churn was increasing, artificially inflated its stock price, and misled investors about the health of its business. The stock dropped over 34% in a single day, falling $6.23 per share from $18.30. That's not a bad quarter. That's a company whose core value proposition is being questioned in federal court. And here's the thing: the churn problem isn't surprising to anyone who has actually used RPA at scale. The bots break. Constantly. One UI update, one field rename, one API change, and your entire automation stack goes dark. A conservative estimate from automation researchers puts the failure rate at 20% per week in a typical 50-bot deployment. That means 10 bots breaking every single week. Someone has to fix those. That someone costs money. A lot of it. A 500K euro RPA implementation costs roughly 75K to 100K euros per year just to keep the lights on. You're not automating your business. You're hiring a maintenance crew for your automation.

Why RPA Was Always a Workaround, Not a Solution

• ●RPA bots follow rigid, pre-programmed rules. Change anything in the UI or the underlying system and the bot fails silently or loudly. Either way, it fails.
• ●The average enterprise RPA deployment requires specialized developers, a dedicated maintenance team, and months of process documentation before a single task is automated.
• ●15 to 20% of your initial RPA investment goes to maintenance every single year. A $1.4M three-year RPA deployment is not an asset. It's a liability that invoices you quarterly.
• ●Over 40% of workers still spend at least a quarter of their work week on manual, repetitive tasks, meaning RPA, even at its best, barely moved the needle for most companies.
• ●UiPath's own annual report lists 'failure of this platform' as a key risk to its entire revenue base, because the company derives substantially all of its revenue from one product that is increasingly being outflanked by AI.
• ●Reddit's RPA community, once a cheerleading section for UiPath, is now full of threads titled 'RPA is dead' and 'Is it worth learning UiPath in 2025?' The practitioners who built these systems are the first ones walking away.

AI Agents Don't Break When the UI Changes. That's the Whole Point.

Here's what makes a computer use AI agent fundamentally different from an RPA bot. An RPA bot sees a button called 'Submit Invoice' and clicks coordinates. You rename that button 'Process Payment' and the bot is dead. A computer use agent sees the screen the way a human does. It reads context. It understands what the interface is trying to do, not just where a pixel is located. That's not a small improvement. That's a completely different category of tool. The benchmark that actually measures this is OSWorld, which tests AI agents on real-world computer tasks across browsers, desktop apps, and terminals. Claude Sonnet 4.5 from Anthropic scored 61.4% on OSWorld. OpenAI's Operator-based approach has similar limitations. These are serious companies with serious models, and they're still leaving nearly 40% of real-world tasks incomplete. The gap between 'impressive demo' and 'reliable production automation' is where most AI computer use tools quietly fail. And then there's the question of whether these tools actually control a real desktop, handle multi-step workflows without handholding, and can run parallel workloads without falling apart. Most can't.

UiPath's Pivot to 'Agentic Automation' Is a Rebrand, Not a Reinvention

To be fair to UiPath, they're not sitting still. Their fiscal year 2025 annual report talks about 'agentic automation' and an 'Enterprise Agent Builder.' They're using the word 'agents' everywhere now. But here's the problem with bolting AI onto a fundamentally brittle architecture: you still have the brittle architecture underneath. UiPath's moat was always the enterprise sales relationships and the massive library of pre-built automations. That moat is real. But it doesn't make the underlying technology less fragile. It just makes it more expensive to replace. Companies that went deep on UiPath are now staring at a choice: keep paying maintenance costs on a system that breaks and a vendor whose stock is in freefall, or rip it out and start over with something that actually works. The Reddit threads from UiPath practitioners in 2025 are brutal. 'Not popular opinion: UiPath and similar tools are dead.' That's not a hot take from an outsider. That's someone who built their career on this stuff and is now telling you to run. When the insiders leave, pay attention.

Why Coasty Exists and Why the Timing Isn't a Coincidence

Coasty was built specifically for the moment we're in right now. Not the moment where AI is a demo. The moment where it has to actually work in production, on real desktops, across real workflows, without a team of consultants holding its hand. On OSWorld, the industry-standard benchmark for computer use AI performance, Coasty scores 82%. That's not a rounding error above the competition. Claude Sonnet 4.5 scores 61.4%. The gap is 20 percentage points, which in production automation translates to the difference between a tool that handles your workflows and a tool that handles your easy workflows and fails quietly on everything else. Coasty controls real desktops, real browsers, and real terminals. Not API wrappers. Not simulated environments. Actual computer use, the way a human would do it, but faster and without the $28,500 annual price tag per head. It runs cloud VMs, supports agent swarms for parallel execution when you need to run the same workflow at scale simultaneously, and has a free tier so you can actually test it before you commit. BYOK is supported if you want to bring your own model keys. The point isn't that Coasty is a cute AI toy. The point is that the thing UiPath promised you in 2019, reliable automation that scales without breaking, is now actually deliverable. Just not by UiPath.