Teachers Are Drowning in Busywork While Schools Debate AI Ethics: A Computer Use Agent Can Fix This Today

The Numbers Are Genuinely Embarrassing

Let's put some real weight on this. Teachers in the US work an average of 54 hours a week according to RAND's 2025 State of the American Teacher survey. Their contracted hours are closer to 40. That gap, those 14 extra hours, is almost entirely administrative. Grading. Paperwork. Data entry into student information systems. Copying grades from one platform to another because the school's LMS doesn't talk to the district's reporting tool. Gallup found in June 2025 that teachers who do use AI weekly are saving an estimated 5.9 hours per week, which adds up to about six full weeks of time per year. Six weeks. And only 30% of teachers are doing it. The other 70% are still doing it manually. Why? Partly because the tools schools have handed them are chatbots, not agents. There's a massive difference between asking ChatGPT to write a rubric and having an actual computer use agent open your grading portal, read student submissions, apply a rubric, populate scores, and send feedback emails. One is a party trick. The other is automation.

What Schools Are Actually Automating (It's Not Enough)

• ●Chatbots that answer FAQ questions on school websites. Genuinely useful for about 4% of the actual workload.
• ●AI writing assistants that help teachers draft lesson plans. Still requires the teacher to sit there and prompt it for 45 minutes.
• ●Plagiarism detectors that flag AI-written student work. Which then creates more work for teachers to investigate and document.
• ●Attendance auto-fill in some districts. One task out of about 200 that need automating.
• ●Grading rubric generators. Again, a chatbot feature, not automation. The teacher still has to do the grading.
• ●Zero schools, as far as I can find, using a real computer use agent to navigate their actual software, execute multi-step workflows, and close tickets without human handholding.

The Cheating Panic Is Eating All the Oxygen

Here's what's infuriating. The entire national conversation about AI in education has been hijacked by the cheating debate. The New York Times ran an opinion piece in August 2025 calling AI cheating a full-blown crisis. Reddit threads with thousands of upvotes declare that AI is ruining education. And sure, students using AI to write essays they didn't think through is a real problem worth discussing. But while administrators are writing 40-page acceptable use policies for students, nobody is asking the more urgent question: why are we not using AI to fix the broken operational infrastructure that's burning teachers out and causing a global shortage? UNESCO confirmed in April 2025 that 90% of annual teacher vacancies come from teachers leaving the profession, not from a pipeline problem. They're leaving because the job is buried under administrative weight. The Learning Policy Institute backed this up the same month. Schools are so focused on whether students should use AI that they forgot to ask whether teachers and administrators should. The answer is yes. Aggressively yes.

What Real Automation Looks Like in a School Setting

Stop thinking about AI in education as a chatbot you talk to. Start thinking about it as a computer use agent that operates software the way a human would, but faster, at any hour, without complaining. Here's what that actually looks like in practice. A computer use agent logs into your student information system, pulls this week's attendance data, cross-references it with the gradebook, flags students who are both absent and falling behind, and drafts outreach emails to parents. All of it. Without you clicking a single thing. Or it opens your district's procurement portal, fills out the supply request form, attaches the required documentation, submits it, and logs the confirmation number in your shared spreadsheet. Or it takes a batch of scanned short-answer responses, applies a scoring rubric you set once, populates grades into the LMS, and generates a class performance summary. These aren't hypothetical future capabilities. This is what computer use AI does right now. The task variety that makes education administration so exhausting, jumping between five different platforms, copying data between systems that don't integrate, filling out the same information in three different formats, is exactly the kind of multi-step, multi-app workflow that a computer use agent is built for.

Why Coasty Is the Right Tool for This

I've looked at what's actually available for this kind of work, and the gap between Coasty and everything else is not subtle. Coasty is the top-ranked computer use agent on OSWorld, the benchmark that actually tests whether an AI can operate real software in real environments, scoring 82%. That's not a marketing number, it's a research benchmark, and no competitor is close. Anthropic's computer use feature and OpenAI's Operator are both real products, but they're general-purpose tools without the infrastructure that institutions need, no agent swarms for parallel execution, no dedicated cloud VMs, no desktop app built for running persistent workflows. For a school district that needs to process 3,000 student progress reports at the end of a grading period, running one agent sequentially isn't good enough. Coasty's swarm architecture runs tasks in parallel, which means what would take a human staff member two days of data entry takes an agent swarm about 20 minutes. It works on real desktops and browsers, not just APIs, which matters enormously in education where the software stack is a chaotic mix of legacy SIS platforms, Google Workspace, Microsoft 365, custom district portals, and whatever the state requires for compliance reporting. None of that plays nicely together. A computer-using AI that can navigate all of it like a human is the only practical solution. There's a free tier to start, BYOK support if your district has API budget constraints, and it doesn't require an IT overhaul to deploy. That last part matters a lot in schools.