AI Prompting Tips

Specificity eliminates guesswork. Each detail — who is reading, why it matters, what to focus on, how long to be — narrows the space of possible answers toward the one you actually need. The poor prompt leaves every one of those decisions to the AI.

Format is part of the prompt. If you define the structure, the output arrives ready to use — not requiring reformatting before you can share it. Asking for a table, a numbered list, short bullets, or a specific word count all shape the output in ways that save time.

This turns the model into a collaborator rather than a guesser. It significantly reduces the chance of receiving a polished but entirely misaligned answer — the most common frustration with AI tools. A few questions upfront saves multiple rounds of revision.

Role and identity shape assumptions. The more the AI understands your specific situation — your constraints, your audience, your responsibilities — the more relevant and actionable the response becomes. The poor prompt generates generic advice; the better prompt generates advice calibrated to your actual job.

If you're unsure how to frame a complex request, ask the AI to help you build the prompt first. This removes the "I don't know where to start" barrier entirely — and the resulting prompt will almost always be more targeted and effective than what you would have written on your own.

Bundling too many requests into a single prompt often produces shallow responses across the board. Sequencing the work gives the model space to think carefully at each stage — and gives you a natural checkpoint to redirect before the next step begins.

When a prompt has many components, AI tools sometimes compress or skip later items to stay concise. Explicitly asking for full coverage prevents this pattern and produces more consistent, complete responses — especially on longer lists or complex multi-part requests.

The first draft is often useful but rarely exceptional. Asking for self-critique and revision in a single prompt produces a substantially stronger answer without requiring a second conversation — and the "skeptical provost" framing sharpens the critique in a direction that's directly relevant to how the work will actually be evaluated.

AI is excellent at generating structured comparisons quickly. Asking for options rather than a single answer gives you better material for decision-making — and often surfaces tradeoffs and considerations you hadn't thought of. The final recommendation also becomes more credible when it emerges from a visible comparison.

AI does not have access to UMD's policies, personnel records, FERPA obligations, or legal context. Use it to organize your thinking and prepare better questions — not to reach conclusions on matters requiring human judgment, institutional knowledge, and proper authority. When in doubt, consult your HR, legal, or compliance office before acting on AI output.

This makes AI output more transparent and easier to verify — especially valuable when stakes are high and the underlying information is incomplete. It also helps you quickly identify which assumptions you need to challenge or confirm before acting on the recommendation.

Not every task warrants the same depth of reasoning. Telling the AI to keep it fast and light on simple drafting tasks saves time. Telling it to think carefully on complex, high-stakes decisions improves quality. Calibrating the effort level to the importance of the task is one of the most underused prompting techniques.