AI Agents at Work: Practitioner
Delegate real multi-step work to AI agents, with the specs, permissions and checks that keep you in control
Put AI agents to work on real, multi-step tasks, and stay in control while they do it. This intermediate course builds on AI Agents at Work: Foundations. You'll write agent specs that hold up, verify what research and analysis agents produce, connect agents to email, calendars, files and business apps with the least access they need, and design approvals, checks and recovery steps that keep people, data and your organization safe. Every lesson uses ETA's ADAPT framework and the ETA Agent Autonomy Ladder, with workplace examples from many sectors and hands-on exercises in the AI tool your organization approves.
What you'll learn
- Write seven-part agent specs: goal, success criteria, inputs, tools allowed, checkpoints, output format and stop conditions
- Verify research agents with a claim ledger and analysis agents with reconciliation tests
- Use browser, computer-use and productivity-suite agents safely, and chain steps into repeatable workflows with gates
- Explain how agents use tools and connectors, including an accurate, plain-language view of the Model Context Protocol (MCP)
- Apply least privilege, map where data goes, and design scheduled and event-triggered workflows that are safe to leave running
- Design approval points that people actually read, and check both an agent's result and its trail
- Recognize the main agent risks using the OWASP Top 10 for LLM Applications, recover from mistakes, and measure value honestly
Who it's for
Working professionals in any sector who have completed AI Agents at Work: Foundations (or have equivalent experience) and are ready to delegate real multi-step work to agents. Useful for operations, finance, HR, marketing, IT, education, healthcare administration, government and small business roles. No coding required.
This course counts toward the ETA AI Empowered Professional: Practitioner credential and the ETA Certified AI Empowered Professional designation. It takes about 2 hours and includes three knowledge checks and a final assessment.