Computer Use

AI that can control your computer to complete tasks

Computer Use represents a fundamental shift in AI capability: from generating text to taking action. Instead of just telling you how to do something, AI can now do it directly by controlling your computer. This is early-stage technology with enormous potential and important considerations.

What Is Computer Use?

Computer Use = AI that can see your screen and operate your computer

The AI can:

  • View what's on your screen
  • Move the mouse and click
  • Type text and use keyboard shortcuts
  • Navigate between applications
  • Complete multi-step workflows

How It Works

  1. You describe a task in natural language
  2. The AI takes screenshots to understand the current state
  3. It plans and executes actions (clicks, typing, navigation)
  4. It verifies results and continues until complete
  5. You review the outcome

Current State of Technology

Available Platforms

| Platform | Feature | Status | Access | | --------- | --------------- | -------- | -------------------------- | | Claude | Computer Use | Beta | API access, Claude Desktop | | OpenAI | Operator | Released | ChatGPT Pro subscription | | Microsoft | Copilot Actions | Preview | Enterprise preview |

What Works Today

Structured, repeatable tasks

  • Filling out forms across systems
  • Data entry from one application to another
  • Navigating familiar interfaces
  • Following documented procedures

Web-based workflows

  • Research tasks with multiple tabs
  • Filling out online forms
  • Extracting information from websites
  • Basic web application operations

What's Still Challenging

  • Complex judgment calls during execution
  • Unpredictable interfaces or pop-ups
  • Tasks requiring real-time adaptation
  • Security-sensitive operations

Use Cases for Education

Administrative Automation

Data entry across systems

Many districts have data that lives in multiple systems without integration. Computer Use can:

  • Pull student information from one system
  • Enter it into another system
  • Repeat across hundreds of records
  • Log what was done for verification

Report generation workflows

When reports require pulling data from multiple sources:

  • Navigate to each data source
  • Extract relevant information
  • Compile into report format
  • Save and organize output

Compliance Documentation

Repetitive documentation tasks

For accreditation, audits, or compliance:

  • Gather evidence from various systems
  • Organize into required formats
  • Fill out standard forms
  • Create audit trails

Potential Future Applications

As the technology matures:

  • Automated attendance reconciliation
  • Cross-system student record updates
  • Bulk communication workflows
  • Routine IT administration tasks

Safety and Supervision

Critical Principles

Always supervise

Computer Use should be monitored, especially for:

  • Tasks involving student data
  • Financial systems
  • Communications sent externally
  • Any action that can't be easily undone

Use sandboxed environments

When possible:

  • Test on non-production systems first
  • Use accounts with limited permissions
  • Run in virtual machines for sensitive tasks
  • Have rollback procedures ready

Verify before completion

Before marking any task "done":

  • Review what actions were taken
  • Verify data accuracy
  • Check for unintended changes
  • Confirm compliance with policies

What NOT to Automate

| Category | Why | | -------------------------- | ------------------------------------------- | | Student discipline records | Requires human judgment, legal implications | | Personnel decisions | Human oversight essential | | Financial approvals | Segregation of duties, accountability | | External communications | Relationship and tone considerations | | Security-sensitive changes | Risk of unintended access |


Getting Started Safely

Recommended Progression

Phase 1: Observation

  • Watch demonstrations of Computer Use
  • Understand what it can and cannot do
  • Identify potential use cases in your work

Phase 2: Low-stakes experimentation

  • Try tasks with no real consequences
  • Use test accounts and sample data
  • Build comfort with the interaction model

Phase 3: Supervised production use

  • Start with low-risk, high-repetition tasks
  • Maintain active supervision
  • Document what works and what doesn't

Phase 4: Broader deployment

  • Expand to additional use cases
  • Develop standard operating procedures
  • Train others on appropriate use

First Tasks to Try

For educators exploring Computer Use:

  1. Research compilation: Have AI research a topic across multiple websites and compile findings
  2. Form filling practice: Fill out a test form from sample data
  3. Navigation practice: Navigate through a complex website to find specific information
  4. Data extraction: Pull information from a website into a structured format

Implications for Education

Administrative Efficiency

The promise: Reclaim hours spent on repetitive data tasks.

Many education administrators spend significant time on tasks that are:

  • Highly repetitive
  • Require moving data between systems
  • Follow predictable patterns
  • Low judgment but high precision

Computer Use can handle these tasks, freeing administrators for higher-value work.

Student Privacy Considerations

When using Computer Use with student data:

  • Ensure AI tools have appropriate data processing agreements
  • Limit access to minimum necessary information
  • Log all actions for accountability
  • Review outputs before any external use
  • Consult with your data privacy officer

Future Workforce Implications

Students entering the workforce will work alongside AI that can take action, not just advise. Understanding this technology helps educators:

  • Prepare students for an AI-augmented workplace
  • Teach appropriate supervision and verification
  • Develop judgment about what to automate
  • Build skills AI cannot replicate

Current Limitations

Technical Constraints

  • Speed: Slower than human experts for familiar tasks
  • Reliability: Can fail on unexpected interfaces
  • Context: Limited understanding of organizational context
  • Cost: Currently expensive for high-volume use

Practical Constraints

  • Access: Requires systems that allow automation
  • Security: May conflict with security policies
  • Accountability: Who's responsible when AI makes errors?
  • Change management: Staff concerns about automation

Action Items

This month:

  • Watch demonstrations of Claude Computer Use or OpenAI Operator
  • Identify 3 repetitive tasks in your work that might benefit
  • Discuss with IT leadership about organizational readiness

This quarter:

  • If available, try Computer Use in a sandboxed environment
  • Develop criteria for what tasks are appropriate to automate
  • Consider policy implications for your organization

This year:

  • Monitor technology development and maturity
  • Plan for gradual integration as reliability improves
  • Prepare staff for working alongside AI automation

Key mindset: This is emerging technology. The goal now is understanding and preparation, not immediate deployment. Experiment safely, learn the capabilities and limitations, and be ready as the technology matures.