Hi Friends and Happy Monday!

Welcome to Week 3 of Monday AI Moments.

Last week we talked about giving AI a role before prompting it. This week, we are talking about the vocabulary you need to know about AI.

Before I share this week’s AI moment, I wanted to touch on something… Learning something new shouldn’t take a huge amount of time or be stressful. I know it can seem that way since we all have hefty workloads. Learning new things isn't something you should have to spend hours doing.  

It’s something you do a little bit at a time, and eventually it sticks. That is why I am doing these bite-sized lessons each Monday that should only take you 5-10 minutes per week – and I hope they help! As long as you are learning or trying something new once a week, you are ahead of the game!

The AI words everyone needs to know

Before we keep building on what AI can do for you, let’s make sure we’re actually speaking the same language. AI conversations throw around a bunch of terms that sound technical but honestly aren’t once you break them down. This week is less about a skill and more about a cheat sheet. Stick this in your back pocket, and every tip after this one will click faster.

Five words worth knowing:

• Prompt (which you already know by now I hope): What you type (or say) to AI. Like texting a very capable assistant who’s never worked for you before.

• Model: The specific AI brain you’re talking to. ChatGPT, Copilot, Claude, Grok, Gemini, etc…are all different models. Each one has strengths and weaknesses, which is why the May 25 email will cover which one to use when.

• Context window: How much AI can remember in one conversation. Think short-term memory. A long document eats it up fast, and once the window fills, earlier stuff starts falling out.

• Hallucination: When AI confidently makes something up that sounds right but isn’t. It’s not lying, it’s guessing. Always double-check names, dates, stats, and anything legal or financial. And also in your prompt, tell it NOT TO hallucinate or let you know when it isn’t 100% confident of something, or tell it to cite its sources.

• Token: The tiny chunks AI reads and writes, roughly 3 to 4 characters each. You don’t need to stress about this one. Just know that long document equals lots of tokens equals slower response.

Why it matters: Most people get stuck with AI because they don’t know what the tool is actually doing under the hood. Once these words click, you’ll stop feeling like you’re throwing spaghetti at the wall and start feeling like you’re in the driver’s seat.

Try this today: Next time you use ChatGPT or Microsoft Copilot and something feels off, ask yourself three quick questions. Did I give it enough context? Is it hallucinating? Is my prompt clear? You’ve got the vocabulary now. Use it like a troubleshooting checklist.

PRO TIP: You can always ask the AI tool you are using if it is hallucinating. See how it responds.

Prompt to use this week:

You are an AI assistant answering questions for a user who cares about accuracy over creativity.

Follow these rules strictly:

1) Ground your answers

- Use only the sources, documents, or data I provide, or clearly established general facts.

- If I attach or paste content, treat it as the primary source and base your answer on that material.

2) Be honest about uncertainty

- If you are not sure, say “I do not know” or “I do not have enough information” and stop.

- Do not guess names, dates, citations, URLs, or statistics.

- If the question is impossible to answer with the information you have, explain what is missing.

3) Label speculation

- If you need to infer or speculate, clearly label that part as “speculation” and keep it short.

- Do not state speculation as fact.

4) No invented sources

- Do not invent books, articles, studies, cases, or links.

- Only mention sources if they are real and you are confident they exist.

5) Always reflect back

At the end of your answer, include a short “Accuracy check” section that:

- Lists any parts of your answer that might be uncertain or approximate.

- Suggests one simple way I could verify the key facts.

Task

[Now respond to this request: “...” using the rules above. Tell is what you want it to do here]

Why this helps

- Clear instructions about grounding and “do not guess” reduce hallucinations by narrowing what the model is allowed to do

- Asking the model to label speculation and call out uncertainty forces it to separate facts from guesses

- Telling the model to work only from your documents or specific sources imitates how tools like NotebookLM and Microsoft Copilot reduce hallucinations by grounding answers in trusted content.

USEFUL RESOURCES

• MIT Sloan: Generative AI Basics – Glossary


• Coursera: Artificial Intelligence Glossary

• Stanford HAI: Brief Definitions of Key Terms in AI
  
  
  See you next Monday!
  
  Lauren, Lulu and Zoey
  
  Barktobank.com
  
  

HAVE AN IDEA FOR THE NEXT MONDAY AI MOMENT, or curious about how to do something with AI? Send us an email!