Five hands-on activities
Build a personal CompTIA tutor bot, a job-search automation, a help desk auto-triage flow, a browser research agent, and put any one of them through a four-test evaluation.
Open activitiesUrban League of Louisiana · Workshop 3 of 3 · Agentic AI
AI that takes actions, not just writes them.
A four-hour workshop on agents, automations, and custom bots. We combine the career skills from Workshop 1 and the tech skills from Workshop 2 into systems that actually do things — with you in the loop on what matters.
Workshop 3 builds on Workshops 1 and 2. You should be comfortable with multi-step prompts and the help desk and CompTIA scenarios from Workshop 2. If you have not taken those, the career activities and tech activities are the prerequisite.
Six worked use cases
Weekly study coach, phishing classifier, LinkedIn outreach drafter, daily ticket digest, flashcard auto-grader, and a candid framework for when paid plans earn the cost.
Open use casesWhat you leave with
- A custom CompTIA tutor bot you can use nightly for the next three months.
- A working job-search automation drafted in Zapier or Make.
- A help desk auto-triage flow tested against a sandbox.
- A safety review of a real browser-agent run on a real research task.
- One documented failure-mode card for an agent you built.
- A clear-eyed framework for when to pay for a tool and when to rotate free tiers.
Plan the day
The four-hour agenda lives at /ws3-agentic/session-3. The live page has a shared timer.
Take-home reading and links: /ws3-agentic/resources.
The honest framing
Most "AI agent" demos in 2026 work 70% of the time. 70% is unacceptable for any task where the wrong action has cost. The skill in this workshop is not building agents that work — it is building agents that fail safely, with humans in the right places.
Every activity in Workshop 3 includes a guardrail step. Take it seriously. The eval discipline is what separates a useful tool from a confidently-wrong tool.