Human Orchestrated Tools and Agents

Most AI tools answer one question at a time. In this 3 hour workshop you will design a multi-step research system — a reusable workflow that helps you find, evaluate, and apply information across the real-world projects you are already working on. By the end of the session you will have a working first draft of your own.

What You’ll Learn

• The difference between asking AI a question and designing a research system — and why the second one compounds the more you use it
• How AI inherits the biases of the web it was trained on — and what to do about it in your own workflow
• A plain-English framework for judging credibility — recency, authority, evidence, relevance
• The vocabulary you actually need: just-in-time help, multi-step AI helpers, triangulation, automation bias
• How to wire credibility checks into your workflow so you trust what comes out

What You’ll Build in the Hands-on Lab:

• Pick one real research problem you actually have — a topic you track, a brief you write, a class of question you answer over and over
• Map it into a 3–5 step research system: what to pull, where to cross-check, how to flag uncertainty
• Test it with a small data set you bring — your PDFs, URLs, notes
• Leave with an MVP workflow and a roadmap for taking it further

What You’ll Leave With

• a tool agnostic contextual documentation that will allow one to continue learning/developing reliable human + ai research capability no matter what tool prevails
• one defined research “problem” unique to personal context/needs with 90 day roadmap on how to increase effectiveness and productivity
• education on how to context engineer foundations for multi-agent systems beyond the session
• AAIA help resources and contacts for system development best practices (people to reach out to when uncertainty/ambiguity becomes overwhelming)

We will use Austin AI Alliance’s own live research program as one teaching case throughout — but everything you build is yours, on your own topic.

Who It’s For

• Folks who have used Claude, ChatGPT, Gemini, or Perplexity and want to go further
• Analysts, researchers, and consultants who do recurring intelligence work
• Civic, economic-development, and policy professionals who need defensible source trails

If you have used one LLM tool, you are qualified.

Before You Arrive

Required:

• Laptop (any modern OS; nothing to install)
• Active accounts on at least two of the following: Claude, ChatGPT, Gemini, Perplexity (free tiers are fine)
• One real research problem or recurring task you would like to work on live
• A small folder of unstructured material — 3–10 PDFs, URLs, or text files — related to that problem

Format