UiPath Lost 34% of Its Stock Value in a Single Day. Maybe RPA Was Always the Problem.

The RPA Business Model Was Always a Trap

Here's how the RPA sales pitch worked. A vendor shows up, promises to automate your repetitive work, charges you $10,000 or more per unattended bot license per year, and then sends in a team of consultants to build brittle scripts that map pixel coordinates and simulate mouse clicks. It works great, right up until the vendor you're scraping updates their UI, your IT team patches a system, or someone resizes a browser window. Then the bot silently fails, your team doesn't notice for three days, and you're paying a developer to go fix it. Reddit threads from UiPath's own community are full of this. One developer summed it up bluntly: the maintenance overhead is a full-time job. You didn't automate your work. You automated the creation of new work. Gartner estimates that over 40% of agentic AI projects will be canceled by end of 2027, largely because companies are confusing legacy RPA platforms with actual intelligent automation. UiPath knows this is a problem, which is why they've been desperately rebranding everything as 'agentic automation' since 2024. Slapping the word 'agentic' on a bot that still breaks when a dropdown menu changes isn't a product evolution. It's marketing.

What RPA Actually Costs You (The Numbers They Don't Put in the Deck)

• ●UiPath unattended bot licenses run $10,000+ per bot per year at enterprise scale, before consulting, implementation, or maintenance fees
• ●RPA maintenance typically consumes 30-50% of the original implementation budget every single year, according to industry benchmarks
• ●A single broken bot can take days to diagnose and repair, and most organizations run dozens of bots across critical workflows
• ●UiPath's own stock fell 34% in one day in May 2024 after issuing weak forward guidance, wiping billions in market cap and triggering a securities fraud class action lawsuit
• ●The class action complaint alleged that UiPath's internal meetings in 2023 and 2024 revealed awareness of serious product and growth problems that weren't disclosed to investors
• ●Duke University research found nearly 60% of businesses have implemented automation, yet most report significant gaps between automation promises and actual outcomes
• ●Companies report spending more time managing bot failures than they saved by deploying the bots in the first place

What a Real Computer Use Agent Actually Does

A computer use agent doesn't follow a script. It looks at a screen, the same way you do, reads what's there, and decides what to do next. If the button moved, it finds the button. If the form added a new field, it fills it. If the website redesigned its checkout flow overnight, the agent figures it out. This is not magic. It's just a fundamentally different architecture. RPA bots are like GPS directions that only work if every road is exactly where it was when the directions were written. Computer-using AI is like a driver who can read signs, adapt to detours, and still get you there. The OSWorld benchmark is the industry's most rigorous test of how well an AI agent can actually operate a real computer, handling real software, in real conditions. Claude Sonnet 4.5 scores 61.4% on OSWorld. That's impressive. Coasty scores 82%. Nobody else is close. That gap isn't a rounding error. It's the difference between an agent that handles your most complex, unpredictable workflows and one that gets stuck when a modal popup appears.

UiPath's Desperate Pivot and Why It Won't Work

To be fair to UiPath, they see the threat. Their fiscal year 2025 annual report talks extensively about 'agentic orchestration' and integrating AI agents into their platform. Their fall 2024 event was literally titled 'reframing enterprise AI with controlled agency.' But here's the problem. You can't bolt intelligence onto a foundation built for rigid rule-following. UiPath's core architecture is still RPA. The bots still need to be built, maintained, and babysat by humans. The 'AI' layer they're adding is mostly a wrapper around existing models, not a rethink of how automation works. It's like a taxi company adding an app and calling themselves a rideshare platform. The underlying economics haven't changed. You're still paying for the car, the driver, the maintenance, and now also the app subscription. Meanwhile, companies using native computer use agents are deploying automation without writing a single script, without a dedicated RPA developer on staff, and without a support contract that costs more than the original license. The Reddit community for UiPath developers is already asking hard questions about where their careers go as AI agents eat the work they used to get paid to do. That's not a community that feels like it's riding a growth wave.

Why Coasty Exists

I'm not going to pretend I stumbled onto Coasty by accident. I was looking for something that could handle the kind of messy, multi-step computer work that RPA was always supposed to automate but never really did. Cross-application workflows. Tasks that require reading context, making judgment calls, and recovering from unexpected states. Coasty is a computer use agent that controls real desktops, real browsers, and real terminals. Not API integrations. Not Zapier triggers. Actual computer use, the way a human does it. It scored 82% on OSWorld, which is the benchmark that actually matters for this kind of work. For context, that's higher than every competitor, including the models from Anthropic and OpenAI that have been getting all the press. You can run it as a desktop app, spin up cloud VMs, or run agent swarms for parallel execution when you need to process work at scale. There's a free tier if you want to actually test it before you commit to anything, and BYOK support if you have your own API keys. The comparison to UiPath isn't even close once you run the same workflow through both. One requires a developer, a license, and a maintenance contract. The other requires describing what you want done. That's the whole pitch, and it holds up.