Monday field guide
A Plain-Language Framework for What Generative AI Can and Cannot Do
A simple way to think about generative AI is to ask what it is best at, where it struggles, and what still needs human judgment. This framework helps beginners use AI more calmly and carefully.
Dr. Mira Vale is our resident AI expert.
Generative AI can feel mysterious because it often sounds confident, moves quickly, and handles language in a very human-like way. That can make it easy to overtrust it or, on the other side, dismiss it entirely. A calmer approach is to use a simple framework: what it can do, what it cannot do, and what still belongs to you.
This matters because generative AI is not one single skill in a box. It is a pattern-making tool. It produces likely-sounding outputs based on patterns in data and prompts, but it does not automatically understand your goals, your context, or the consequences of a mistake. If you keep that distinction in mind, it becomes much easier to decide when AI is useful and when human judgment should lead.
Start with three questions
When you are deciding whether to use generative AI for a task, ask:
- Is this a language and pattern task?
- Is accuracy or context critical?
- Would a mistake be easy or hard to catch?
These three questions give you a practical filter.
If the task is mainly about drafting, summarizing, rephrasing, organizing ideas, or generating options, AI may help. If the task depends on exact facts, deep context, ethical judgment, or high-stakes decisions, AI should be treated much more cautiously. And if a mistake would be hard to notice, you need extra verification or a different workflow.
That simple test is often more useful than asking, “Can AI do this?” The better question is, “What part of this task is AI good at, and what part should stay with a human?”
What generative AI is good at
Generative AI tends to do well with tasks that involve language, structure, and variation.
It can often help you:
- Draft a first version of an email, outline, or summary
- Rephrase text in a different tone
- Brainstorm ideas or alternatives
- Turn rough notes into a more organized draft
- Explain familiar concepts in simpler language
- Compare options when the criteria are clear
A useful way to think about this is that AI is often strongest when the goal is to create a starting point rather than to finish the job perfectly. It can reduce blank-page friction and speed up repetitive language work.
That does not mean the output is automatically reliable. It means the output may be useful as a draft, a sketch, or a set of options.
What generative AI cannot do well on its own
Generative AI has important limits. It is not a substitute for understanding, responsibility, or real-world verification.
Here are a few things it cannot reliably do on its own:
- Know the truth by itself
- See the full situation unless you provide it
- Guarantee accuracy in every answer
- Make accountable decisions for you
- Understand values and tradeoffs the way a person does
- Notice missing information unless prompted to look for it
It may produce an answer that sounds polished even when it is incomplete, wrong, or too generic. That is why confidence is not the same as correctness. A smooth response is not proof.
This is one of the most important parts of AI literacy: the tool can generate text, but you still need to evaluate meaning, fit, and risk.
A practical framework: four layers of judgment
A beginner-friendly way to understand generative AI is to separate the task into four layers.
1. Generation
This is where AI is most helpful. It creates a draft, list, explanation, or version of something.
Example uses:
- First drafts
- Rewriting for clarity
- Brainstorming headlines or outlines
- Summarizing a long note into key points
2. Verification
This is where you check whether the output is actually correct, complete, and appropriate.
Ask:
- Are the facts right?
- Did it miss anything important?
- Does it match the source material?
- Is it using the right tone and terms?
3. Interpretation
This is where you decide what the information means in your real context.
Ask:
- Does this fit the situation I care about?
- What assumptions are hidden here?
- What would a human expert or stakeholder say?
4. Responsibility
This is the human layer. It is where you decide whether to use, revise, or reject the output.
Ask:
- Should I trust this enough to act on it?
- What would happen if it is wrong?
- Do I need another source, another person, or another process?
The point is not to make AI “better” than it is. The point is to place it in the right part of the workflow.
A hypothetical example: planning a team update
Imagine you need to send a weekly project update to your team. You have a list of notes, but they are messy and scattered.
You ask a generative AI tool to turn the notes into a clean draft with three sections: progress, blockers, and next steps. The tool does a decent job. It organizes the content, removes repetition, and gives you a readable first version.
Now the human work begins.
You check whether the draft:
- Leaves out any important blocker
- Uses the correct project names and dates
- Reflects the latest status
- Sounds respectful and clear for your team
You then revise the draft, delete one inaccurate sentence, and add a missing decision point that the AI did not include.
In this example, AI did what it is often good at: structuring language. You did what only a person could do well: verify context, judge importance, and finalize the message.
That is the healthy pattern. AI helps you move faster, but it does not replace your accountability.
A simple checklist for using generative AI responsibly
Before you rely on an output, run through this checklist:
- What is the task? Is it drafting, summarizing, brainstorming, or decision-making?
- What matters most? Speed, clarity, accuracy, nuance, or safety?
- What does the model not know? Did I give enough context?
- What could be wrong? Are there facts, names, numbers, or claims to verify?
- How hard would an error be to catch?
- Who is responsible for the final result?
- Do I need a second review?
If you want a shorter version, use this rule:
Use AI for first drafts and options; use human judgment for final decisions and high-stakes work.
Common mistakes beginners make
A few patterns show up again and again:
Treating a fluent answer as a correct one
Polished writing can hide weak reasoning or missing facts.
Asking a vague prompt and expecting a precise result
If the input is vague, the output will often be vague too. Clear goals and examples improve usefulness.
Using AI where context matters most
A model may not know the unwritten details that shape your real situation.
Not checking the output against a trusted source
Verification is not optional when accuracy matters.
Assuming AI understands your intent
It does not read your mind. It follows patterns in the prompt and the information available.
Letting AI do the final thinking
The tool can support thinking, but it should not replace it.
These mistakes are normal when someone is new. The goal is not perfection. The goal is to get more aware of where the tool fits.
The real strength of generative AI: support, not certainty
A useful mindset is to see generative AI as a support tool for language work, not a source of guaranteed truth.
That framing keeps expectations realistic. It helps you notice when AI is saving time, when it is creating a useful draft, and when it is simply not the right tool.
It also makes experimentation safer. You can try small, low-risk tasks first: rewriting an email, summarizing meeting notes, or brainstorming a table of contents. Then you can gradually learn where the tool helps you most.
A realistic next step
The best next step is a small experiment.
Choose one everyday task that is low-stakes and easy to verify. Ask a generative AI tool for a draft, then compare its output with your own version. Notice:
- What it did well
- What it missed
- What it misunderstood
- What you had to correct
That exercise will teach you more than a long list of abstract rules. Over time, you will build a practical sense of where generative AI is genuinely useful and where your own judgment should stay in charge.
That is the heart of AI literacy: not fear, not hype, but clear eyes and careful use.
Key takeaways
- Generative AI is strongest at drafting, summarizing, rephrasing, and brainstorming.
- A polished answer is not the same as a correct answer.
- Use a four-layer frame: generation, verification, interpretation, and responsibility.
- High-stakes or context-heavy tasks need extra human judgment and checking.
- AI is best treated as a support tool, not a source of certainty.
- Small experiments are a practical way to learn where AI helps and where it falls short.