You can spend a long time collecting prompts without changing the way you work. A practical learning session starts with an outcome: a brief to write, a page to improve, a research question to answer, or a repeated task to simplify.
Choose a manageable piece of work
Pick something small enough to review from beginning to end. You should know what a good result looks like, or have someone who can help define that standard. Use non-sensitive material while you are learning.
Write down the audience, purpose, constraints, and output format. This becomes your task brief. The brief is useful across tools because it explains the work instead of relying on a magic phrase.
Practice the decisions
Ask for a first attempt, then inspect it. What is accurate? What is generic? What is missing? Which parts need your judgment? Make a deliberate revision and notice whether the result improves.
Learning is not just getting the tool to produce something. It is learning when to provide more context, when to ask a better question, and when to reject an answer that does not meet the standard.
Match the session to your goal
You might choose Claude for a writing or thinking exercise, a Claude workflow session for a repeated process, Claude Design for a visual concept, or ChatGPT for a business brief or analysis. Our training pages describe the work we can practice together and what to bring.
For ChatGPT Work, OpenAI describes workflows that use files and available tools to produce reviewable work, with GPT-6 Astra available for demanding tasks depending on access. See OpenAI’s official guide. Product availability and account features can change; check access before a session.
Leave with a repeatable practice
Save the brief, a good example, and a short quality checklist. Run the task again with different inputs. That second attempt tells you whether you learned a process or simply got a helpful result once.
If you are learning with a team, compare how people judged the output. A shared standard can be as valuable as a shared tool.
Your next move
Choose one task from this week’s workload. Bring it, along with a non-sensitive example, to a practical AI working session.