Get in Touch
 Duration 7 hours

Course Outline

  • Day 1: Delegation of workflows via browser agents
    • Mechanics of browser agent operations
      • Page analysis, action planning, interaction execution, and verification
      • Distinctions from conversational research methodologies
      • Current reliability benchmarks and operational limitations
      • Practical exercise: Monitor an agent executing a basic workflow and document each procedural step.
    • Comparative analysis of available tools
      • ChatGPT Agent and the Atlas browser
      • Claude integration within the Chrome ecosystem
      • Perplexity Comet capabilities
      • Evaluation of availability, cost structures, and data security protocols
      • Practical exercise: Execute an identical research objective using two distinct platforms and analyze the variances.
    • Research and comparative assessments
      • Systematic comparison of vendors, pricing, and terms across multiple sources
      • Generation of tabular results with source citations
      • Verification of data accuracy against primary sources
      • Practical exercise: Generate a vendor comparison matrix based on five criteria and validate two specific data points.
    • Form processing and web application integration
      • Population of forms using pre-existing datasets
      • Extraction of data from web applications into structured tables
      • Implementation of mandatory checkpoints prior to submission, payment, or deletion
      • Practical exercise: Populate a test form with data from a table and conduct a pre-submission audit.
    • Security considerations
      • Mitigation of prompt injection via hidden instructions on web pages
      • Management of authenticated sessions, credentials, and payment information
      • Utilization of isolated browser profiles and approval workflows
      • Practical exercise: Interact with a prepared test page containing hidden instructions and assess the agent’s response.
    • Governance and usage protocols
      • Definition of permitted task categories
      • Establishment of review and approval hierarchies
      • Planning for recurring automated tasks
      • Practical exercise: Draft a one-page policy document and three specific use cases for internal teams.
    • Outcome: Each participant will have delegated two tasks to a browser agent, verified the outcomes, and established team usage guidelines for government contexts.

Requirements

  • Laptop equipped with Chrome, an active account for at least one supported tool (ChatGPT, Claude, or Perplexity), and a non-sensitive test account for exercises.
  • Programming expertise is not required.

Target Audience

  • Specialist staff, administrative assistants, and researchers seeking to delegate recurring web-based and application-centric tasks to AI agents for government operations.

Number of participants


Price per participant

Testimonials (3)

Upcoming Courses

Related Categories