Humanity’s Last

Prompt Engineering
Guide.
Most guides on prompt engineering either overcomplicate the basics or give you vague
tips that don’t hold up when you actually try to use them.

This one’s different.

We’ve done the homework for you — gathering best practices from OpenAI, Google,
Anthropic, and real-world testing — and broken it all down into something you can
actually use.

Whether you’re in sales, operations, marketing, or leadership, this guide will show you
how to get dramatically better results from ChatGPT, Claude, Gemini, and other
chat-based AI tools.

How to Use This Guide

This guide is designed to help you become dramatically more effective at using AI tools
like ChatGPT, Claude, and Gemini — whether you’re just getting started or already using
them in your daily work.

Here’s what to expect:

Forward Future: Learn faster. Work smarter. 1

🧭 Structured for Action
●​ Section 1 introduces how prompting works and why it matters.
●​ Section 2 explains what makes a prompt work — and how models think.
●​ Section 3 helps you diagnose and fix bad prompts fast.
●​ Section 4 breaks down 11 foundational prompting techniques with real
examples.
●​ Section 5 gives ready-to-use templates for common business roles.
●​ Section 6 includes a scorecard and worksheet to refine and evaluate your
prompts.
●​ Section 7 offers a glossary of key terms for beginners and pros alike.​

💡 How to Use It
●​ Skim the techniques — then go deep on the ones most relevant to your
workflows.
●​ Use the diagnostic section (Section 3) any time the AI gives poor or vague
results.
●​ Return to the scorecard when refining a prompt for better consistency or results.​

👥 Who It’s For​

This guide is for professionals — in marketing, sales, operations, leadership, and beyond
— who want to get better results from AI without needing technical expertise.

What You'll Learn

●​ How prompts actually work behind the scenes

●​ Proven techniques to improve accuracy and creativity
●​ Ready-to-use templates for common business roles
●​ Common mistakes to avoid

Forward Future: Learn faster. Work smarter. 2

Section 1: Introduction to Prompt Engineering

Prompting isn’t about tricking AI — it’s about teaching it to think clearly. Large language
models (LLMs) like ChatGPT, Claude, and Gemini are fundamentally prediction engines.
Given an input (your prompt), they generate the most likely next word based on patterns
learned from massive amounts of text. That makes your prompt the blueprint.

A vague request will produce vague output. But a clear, well-structured prompt? That’s
how you get useful, specific, and often surprisingly intelligent responses.

Why Prompt Engineering Matters

Prompt engineering is the skill of designing effective instructions for AI. It doesn’t
require coding, ML knowledge, or advanced technical skills. What it does require is the
ability to be clear, specific, and intentional with your inputs.

A good prompt tells the model: who it is (“You are a product strategist…”), what to do
(“Summarize this in 3 bullet points.”), what it’s working with (Input: text, table, scenario,
etc.), and how to respond (e.g., bullet list, JSON, tone, word count).

Once you learn how to structure that input — and iterate when things go sideways — you
can: summarize complex documents in seconds, brainstorm new product ideas or
messaging, analyze data, extract patterns, or generate insights, role-play expert
conversations, and automate repetitive writing and analysis tasks.

Forward Future: Learn faster. Work smarter. 3

Prompting Is a Leverage Skill

The better you prompt, the more productive and valuable AI becomes in your workflows.
It’s not just a shortcut — it’s a force multiplier.

You don’t need to become a “prompt hacker.” You just need to learn how to speak the
model’s language. This guide will show you how.

That’s why prompt engineering is one of the most valuable skills in the AI era.

Forward Future: Learn faster. Work smarter. 4

Section 2: What Makes a Prompt Work?

Think of it like giving a highly capable assistant a task — if you don’t explain what you
want, how you want it, and why it matters, the results may be vague, verbose, or just
flat-out wrong.

Prompts are affected by more than just what you type. These factors all play a role:

Factor Why It Matters

Model Each LLM has unique strengths, capabilities, and quirks

Context Input quality (documents, examples) impacts reasoning and accuracy

Structure Clear formatting improves output consistency and usefulness

Style + Tone You can control formality, voice, or persona

Model Output length, temperature, and sampling all influence creativity vs.
Settings precision

Prompt engineering is not a one-and-done process — it’s iterative. You’ll often need to
test, tweak, and refine to get the best results.

Behind the Scenes: How Language Models Think

LLMs are prediction engines. They don’t think like humans — they generate the next
most likely word (token) based on the sequence you give them, plus everything they’ve
seen during training.

Forward Future: Learn faster. Work smarter. 5

That means your prompt is effectively setting the stage: you’re steering the model’s
attention, telling it what type of answer to aim for, and how to structure its response.
Small changes in how you phrase your prompt can lead to radically different outputs.

LLM Output Configuration (What You Can Control)

Most AI platforms (like ChatGPT, Claude, Gemini) let you adjust settings that affect how
responses are generated:

Setting What It Does

Temperature Controls randomness. Lower = more focused, higher = more creative.

Max tokens Caps the length of the response. Prevents run-ons

Top-p / Top-k Controls which words the model can choose from when generating
responses.

Forward Future: Learn faster. Work smarter. 6

Section 3: Fixing Bad Prompts — Common Mistakes &
How to Improve Them
Poor results often come from unclear, vague, or under-specified prompts. Below are
common weak prompts and their improved versions — showing exactly how to fix them
for better structure, clarity, and performance.

Problem ❌ Weak Prompt ✅ Improved Prompt

Too vague Write a summary. Summarize the article below in 3 bullet points. Focus
on key findings, avoid repeating the introduction.

No audience Rewrite this for Rewrite this for a busy executive audience. Use short
specified clarity. sentences and strip out nonessential background.

Missing Help me with this You are a brand copywriter. Improve the tone of this
role/context draft. draft to make it more confident and modern.

No format What's a good Suggest 3 alternatives in a numbered list. Include

instruction alternative? 1–2 sentence explanations for each.

No reasoning What’s the best Evaluate these 3 options. List pros and cons for
requested option here? each, then recommend one with a short rationale.

Unclear intent Turn this into a Convert the following into a 3-line Slack message
message. that is clear, friendly, and includes a call to action.

Lacks structure Help me improve Rewrite this performance review to follow this
this. structure: achievements, challenges, and next steps.

Forward Future: Learn faster. Work smarter. 7

Use this as a quick diagnostic when a model gives poor output:

1.​ Am I being too vague?​

Be specific about the task and expectations.​

2.​ Did I include a role or point of view?​

Adding “You are a…” sets the tone and mindset.​

3.​ Is the input complete and relevant?​

Include all necessary information for the model to reason effectively.​

4.​ Have I requested a clear format?​

Specify if you want bullets, a paragraph, JSON, etc.​

5.​ Am I asking for reasoning?​

If judgment is involved, ask the model to “think step by step” or explain its logic.​

6.​ Have I broken the task into smaller parts if needed?​

Split complex requests into multiple, focused steps.​

Forward Future: Learn faster. Work smarter. 8

Section 4: Prompting Techniques

Prompting is a skillset. These eleven foundational techniques will help you structure
clearer instructions, improve model performance, and adapt to different use cases
across business and technical workflows.

Each technique below includes a description, when to use it, and a simple example.

1. Zero-Shot Prompting

What it is: A single instruction without any examples.​

When to use it: For simple or well-understood tasks where context is obvious.

More Examples:

●​ Summarize the following transcript into a one-paragraph executive briefing suitable

for a board meeting.
●​ Generate 3 headline options for a homepage that emphasizes speed, simplicity, and
trust.
●​ Based on the meeting notes below, write an email recap with next steps and
responsible owners.
●​ Extract and rank the top 3 objections raised in this sales call transcript.

Forward Future: Learn faster. Work smarter. 9

 Tips:
🟢 Beginner: Start with simple, clear instructions using action verbs like
“Summarize,” “List,” or “Write.” Avoid vague words like “help” or “explain.”

🔵 Intermediate: Add a format instruction (e.g., “in 3 bullet points” or “as a table”)
to shape the output and reduce ambiguity.

⚫ Advanced: Combine multiple directives in a single sentence to pack precision

(e.g., “Summarize in 3 bullets, each under 10 words, focusing on customer pain
points.”)

2. Few-Shot Prompting

What it is: You provide one or more examples to guide the model.​
When to use it: When the task involves structure, tone, or formatting that the model
might not infer on its own.

Forward Future: Learn faster. Work smarter. 10

More Examples:

●​ Example: I want a large pizza with mushrooms and basil. JSON: {"size": "large",
"toppings": ["mushrooms", "basil"]}. Now convert this: I’d like a small pizza with
olives and tomato.
●​ Here is an ideal product changelog entry: Improved onboarding flow: Faster load
times, clearer CTAs, and better mobile UX. Now write one for the feature update
described below.
●​ Here are two customer support responses that defuse frustration with empathy and
clarity. Write a third using the same tone and structure.
●​ Based on these examples of investor update summaries, write one for this month’s
company performance.
●​ Here are a few ideal Slack status messages for when someone is heads-down.
Create 5 more in the same style.​

Tips:
🟢 Beginner: Copy the structure of one example and change the inputs. This
helps you build pattern recognition.

🔵 Intermediate: Provide a range of examples (short/long, formal/casual) to help

the model generalize the style or logic.

⚫ Advanced: Fine-tune examples to emphasize edge cases, errors, or style

nuances. Use deliberately contrasting examples to teach distinctions.

3. System Prompting

What it is: Set clear rules for how the model should behave or structure output.​
When to use it: To control tone, format, length, or behavior from the beginning.

Forward Future: Learn faster. Work smarter. 11

 ​

More Examples:

●​ Always answer using the STAR method (Situation, Task, Action, Result). Respond
only in bullet points.
●​ Only use language that a 10th-grade reader can understand. Avoid technical terms
and acronyms.
●​ Provide your answer in Markdown format, using headings, subheadings, and bolded
key terms.
●​ Every time you respond, include a short summary sentence first, followed by your
full explanation.​

Tips:
🟢 Beginner: Use simple system rules like “Respond in a friendly tone” or
“Answer in bullet points.”

🔵 Intermediate: Layer constraints like tone, format, and audience all in one
system prompt.

⚫ Advanced: Use system prompts to enforce behavior across multiple turns in a

chat (e.g., always speak as a mentor, always return Markdown with code blocks,
etc.).

Forward Future: Learn faster. Work smarter. 12

4. Role Prompting

What it is: Assign a persona or point of view.​

When to use it: When the model needs to adopt a specific tone, expertise level, or
perspective.

More Examples:

●​ You are a financial advisor. Recommend three strategies for someone with $10,000
to invest and a five-year time horizon.
●​ You are a senior PMM at a B2B SaaS company. Write a 3-sentence value
proposition for a new AI workflow feature.
●​ You are a startup CEO preparing for a pre-seed pitch. Reframe this vision statement
to be more investor-friendly.
●​ You are a customer success manager. Draft a proactive check-in email for an
enterprise client showing early signs of churn.
●​ You are a technical recruiter. Review the résumé snippet below and write a 2-line
summary of the candidate’s profile.

Forward Future: Learn faster. Work smarter. 13

 Tips:
🟢 Beginner: Start with “You are a…” followed by a familiar job title (e.g., “You are
a content strategist”).

🔵 Intermediate: Match the role to the task. Use domain-specific roles (e.g., “You
are a data privacy consultant”) to shape tone and accuracy.

⚫ Advanced: Combine roles with context and format constraints for nuanced
control (e.g., “You are a veteran PMM writing a short memo for legal approval”).

5. Contextual Prompting

What it is: Provide relevant background, data, or a scenario.​

When to use it: When task success depends on domain-specific context or past
conversation history.

Forward Future: Learn faster. Work smarter. 14

More Examples:

●​ You are writing for a retro gaming blog. Suggest three article ideas based on 1980s
arcade culture.
●​ Using the customer persona and product description below, write a 2-sentence ad
hook that appeals to first-time users.
●​ Here’s a summary of our brand’s tone guidelines. Rewrite the following help center
article to match that tone.
●​ Given our revenue goals and Q3 priorities, generate 3 OKR options for the sales
enablement team.
●​ Review this engineering roadmap and suggest a revised timeline, given the team’s
current velocity and bandwidth constraints.​

Tips:
🟢 Beginner: Copy and paste a relevant doc or snippet before the prompt to give
the model something to work with.

🔵 Intermediate: Use summaries instead of raw data when the input is long.
Pre-frame: “Here’s what you need to know…”

⚫ Advanced: Chain multiple inputs (e.g., persona + doc + goals), and explicitly
state how they relate (e.g., “Given this persona and these product notes…”).

6. Step-Back Prompting

What it is: Ask the model to solve a general question first, then apply that insight to a
specific task.​
When to use it: For creative tasks, complex reasoning, or scenario planning.

Forward Future: Learn faster. Work smarter. 15

Prompt Before:

Step-back Version:

​
More Examples:

●​ Before writing the email, list 3 things the recipient likely values based on their role
and company. Then write the email using that context.
●​ First, outline the characteristics of a high-performing user onboarding flow. Then
critique the one below using that framework.
●​ Before drafting product positioning, summarize how our top 3 competitors describe
themselves. Then write our version to differentiate.
●​ Start by identifying the key emotion this campaign should trigger in customers.
Then write a subject line that evokes it.​

Forward Future: Learn faster. Work smarter. 16

 Tips:
🟢 Beginner: Ask a warm-up question before the real task (e.g., “What does the
audience care about?” → “Now write the email”).

🔵 Intermediate: Use it to deconstruct complex decisions or plans (e.g., goals →

actions → message).

⚫ Advanced: Use it for self-discovery or creative ideation chains. Ask the model
to generate principles first, then apply them to the prompt task.

7. Chain-of-Thought Prompting (CoT)

What it is: Instruct the model to show its reasoning step by step.​
When to use it: For math, logic, planning, or anything that benefits from transparent
reasoning.

​
​
More Examples:

●​ Walk through the steps needed to calculate customer acquisition cost (CAC) using
the data provided. Then provide the result.

Forward Future: Learn faster. Work smarter. 17

 ●​ Given this list of product issues, identify which could cause the highest drop in NPS,
and explain your reasoning.
●​ List the steps a support agent should take to de-escalate the situation described in
the chat log.
●​ Break down the decision-making process for whether to renew or sunset the
feature described. Include cost, usage, and team feedback.​
​

Tips:
🟢 Beginner: Add “Explain your reasoning” or “Think step by step” to the end of
your prompt.

🔵 Intermediate: Ask the model to list pros and cons, compare options, or walk
through logic to justify a recommendation.

⚫ Advanced: Combine with structured outputs (e.g., tables of tradeoffs,

multi-step formulas) or nested reasoning paths.

8. Self-Consistency

What it is: Run the same prompt multiple times and choose the most frequent or
consistent result.​
When to use it: For high-stakes or ambiguous tasks where accuracy matters.​
Note: Requires a model that supports variability via temperature (e.g. 0.7 or higher).

Forward Future: Learn faster. Work smarter. 18

 ​

More Examples:

●​ Ask a model to classify an ambiguous email as “important” or “not important”

multiple times, then take the majority vote.
●​ Classify this review as positive, neutral, or negative five times and return the most
frequent label with reasoning.
●​ Generate three different explanations for this customer churn event. Then
summarize the one that appears most consistently.
●​ Write five taglines for this product feature. Then identify the most recurring theme
or hook among them.
●​ Answer this ambiguous question three times, then decide which version is most
likely to align with the user’s intent.​

Tips:
🟢 Beginner: Run the same prompt a few times and compare results. Choose the
best version manually.

🔵 Intermediate: Ask the model to repeat a task 3–5 times and select the most
common or consistent answer.

⚫ Advanced: Instruct the model to generate multiple answers, evaluate them for
consistency or correctness, and select the best one.

Forward Future: Learn faster. Work smarter. 19

9. Tree of Thoughts (ToT)

What it is: Explore multiple reasoning paths instead of a single linear answer.​
When to use it: For brainstorming, decision-making, or tasks with multiple valid
outcomes.​
Note: Often used with agents or frameworks that track branching logic.

​
​
More Examples:

●​ Explore three different GTM strategies for launching this product in Q4. Then
evaluate their tradeoffs and recommend one.
●​ Generate 3 unique frameworks for organizing this internal knowledge base. Justify
each approach and select the most scalable one.
●​ Outline 3 alternative messaging directions for this new feature. Then score them for
clarity, originality, and alignment with brand tone.
●​ Imagine three distinct customer objections during the onboarding process. For
each, propose a mitigation strategy.​

Forward Future: Learn faster. Work smarter. 20

 Tips:
🟢 Beginner: Ask the model for “3 different ideas” before choosing one.
🔵 Intermediate: Prompt for multiple options, then ask the model to evaluate
them for clarity, impact, or feasibility.

⚫ Advanced: Use ToT as a decision-making engine — branch multiple reasoning

paths, evaluate tradeoffs, then synthesize the best choice.

10. ReAct (Reason + Act)

What it is: The model reasons step-by-step and performs actions, such as searching the
web or calling a tool.​
When to use it: For tool-augmented agents or systems that require external data
retrieval or interaction.

​
​
More Examples:

●​ Search how many children each member of Metallica has. Sum the total. Output
only the final number.

Forward Future: Learn faster. Work smarter. 21

 ●​ Search for this company's most recent funding round. Then write a congratulatory
email that references the announcement.
●​ Use a code interpreter to calculate the monthly burn rate from this spreadsheet.
Then write a plain-English summary.
●​ Search for the latest update to the GDPR regulation. Then suggest one change to
our current data policy.
●​ Find the Glassdoor average rating for this company. Based on the result, write an
outreach message addressing potential concerns.

Tips:
🟢 Beginner: Use tools like “Search” or “Calculate” and then manually combine
results with AI output.

🔵 Intermediate: Use prompts like “Search for X, then summarize it in 2 lines” to

simulate a chain of thought + action.

⚫ Advanced: Combine tool use, reasoning, and generation in one loop (e.g.,
“Search X → Compare Y → Generate Z”). Use agents or code interpreter if
available.

11. Automatic Prompt Engineering (APE)

What it is: Use AI to generate and test variations of prompts.​

When to use it: When you’re designing for scale, training a chatbot, or optimizing prompt
performance.

Forward Future: Learn faster. Work smarter. 22

 ​

More Examples:

●​ Write 10 alternate prompts for asking AI to summarize a technical article for a

general audience. Then pick the clearest one.
●​ Generate 5 prompt variations for helping a user reframe negative self-talk. Rank
them from most supportive to most neutral.
●​ Create 7 prompt versions that ask users to describe product feedback. Tag each by
tone (formal, casual, creative).
●​ Produce 5 options for asking ChatGPT to generate campaign ideas for a fintech
company. Rate them based on clarity and specificity.​

Tips:
🟢 Beginner: Ask the model to rewrite your prompt in 3 clearer variations and
choose your favorite.

🔵 Intermediate: Generate 5–10 versions with different tones, formats, or levels

of detail — then test them across tasks.

⚫ Advanced: Set up side-by-side testing or use frameworks like PromptLayer,

DSPy, or PromptFoo to evaluate performance at scale.

Forward Future: Learn faster. Work smarter. 23

Section 5: Role-Based Prompt Templates

These prompts are designed for everyday use across key business functions. Each one
is adaptable — just swap in your input. Tone/style suggestions are included where
applicable.​

Operations
Prompt 1

You are a process improvement analyst. Review the following workflow and suggest
three ways to reduce manual steps.

Tone: Analytical, concise​

Prompt 2

You are an operations manager. Based on the data below, identify any scheduling
conflicts and recommend an optimized shift plan.

Tone: Objective, action-oriented

Prompt 3

You are a logistics coordinator. Rewrite these internal instructions so they’re clear,
step-by-step, and easy to follow for new hires.

Format: Ordered list

Forward Future: Learn faster. Work smarter. 24

Sales
Prompt 1

You are an outbound strategist. Write a LinkedIn message to a VP of Finance at a

mid-sized SaaS company. Make it conversational, relevant, and end with a question.

Tone: Conversational, focused​

Prompt 2

You are a sales coach. Based on this call transcript, identify three missed
opportunities and suggest a better framing of the product’s value.

Tone: Constructive, direct​

Prompt 3

You are a strategic account executive. Write a follow-up note after a promising
discovery call. Include 1 recap line, 2 insights, and 1 CTA.

Format: Structured email outline​

Marketing
Prompt 1

You are a conversion copywriter. Rewrite the following landing page headline to make
it more urgent, specific, and benefit-driven. Provide three options.

Tone: Punchy, persuasive

Forward Future: Learn faster. Work smarter. 25

Prompt 2

You are a content strategist. Turn this blog post into a three-post LinkedIn carousel
script with a hook, a breakdown, and a CTA.

Format: Slide-by-slide copy outline

​
Prompt 3

You are a brand voice expert. Review this homepage and suggest two changes to
make it more consistent with a warm, authoritative tone.

Tone: Editorial​

Management
Prompt 1

You are a leadership coach. A team lead is struggling with prioritization. Share three
frameworks they can use to manage competing requests.

Tone: Supportive, instructive

​
Prompt 2

You are a department head. Draft a message to your team acknowledging recent
tension and outlining next steps to restore alignment.

Tone: Empathetic, confident

Forward Future: Learn faster. Work smarter. 26

Prompt 3

You are an executive preparing for a board update. Summarize your top three metrics,
growth levers, and open risks in under 150 words.

Format: Bulleted or block-style summary

Content Development
Prompt 1

You are a senior editor. Improve the clarity, tone, and structure of this article. Make it
more engaging without losing substance.

Tone: Crisp, confident

​
Prompt 2

You are a content repurposing expert. Turn this long-form article into one tweet
thread, one newsletter excerpt, and one TikTok script.

Format: Multi-output summary

​
Prompt 3

You are a product copywriter. Write three microcopy options for a call-to-action
button encouraging users to upgrade.

Tone: Clear, value-driven​

Forward Future: Learn faster. Work smarter. 27

Data Analysis
Prompt 1

You are a business analyst. Interpret the following product usage report and highlight
three meaningful trends.

Tone: Insightful, plain-language​

Prompt 2

You are a data storyteller. Write a 5-sentence summary of this chart for an executive
audience with no technical background.

Tone: Accessible, high-level​

Prompt 3

You are a customer insights lead. Based on this NPS feedback dataset, extract two
recurring themes and suggest one recommendation.

Format: Bullet points or numbered summary

Learning & Development

Prompt 1

You are an instructional designer. Turn this onboarding checklist into a simple 5-step
training module for new hires.

Format: Step-by-step breakdown

Forward Future: Learn faster. Work smarter. 28

Prompt 2

You are a corporate trainer. Draft a Slack message inviting employees to a new AI
literacy workshop. Make it short, clear, and inviting.

Tone: Friendly, motivating​

Prompt 3

You are a learning coach. Based on this employee feedback, recommend two
personalized learning paths and one potential mentor pairing.

Tone: Supportive, customized

Forward Future: Learn faster. Work smarter. 29

Section 6: Prompt Scorecard & Worksheets

Prompt engineering is an iterative process. The more you test, adjust, and document,
the more consistent your results become. Use the following scorecard and worksheet to
improve prompt quality, troubleshoot issues, and build a reusable prompt library.

Prompt Quality Scorecard

Use this 7-point checklist to evaluate the strength of a prompt. For each item, give
yourself a score from 1 (no) to 5 (excellent).

Question Score (1–5)

1. Is the task clearly defined?

2. Did I assign a clear role or persona to the model?

3. Did I provide the right context or background

information?

4. Did I specify a desired output format (list, paragraph,

JSON)?

5. Did I include instructions for tone, length, or

constraints?

6. If applicable, did I request reasoning or step-by-step

thinking?

7. Is the prompt easy to understand and free of

ambiguity?

Scoring guide:

Forward Future: Learn faster. Work smarter. 30

 ●​ 30–35 = Strong, high-confidence prompt
●​ 20–29 = Decent, may benefit from revisions
●​ Below 20 = Likely to produce inconsistent or unclear outputs

Prompt Refinement Worksheet

Use this template to document and refine your most important prompts over time.

Field Notes

Prompt Name e.g. Weekly Report Summary

Goal e.g. Summarize usage data into executive-facing bullet points

Model e.g. ChatGPT-4, Claude 3, Gemini Pro

Temperature e.g. 0.2 (factual), 0.8 (creative)

Output Format e.g. JSON, table, three bullet points

Prompt (Initial) [Paste full prompt]

Output Sample [Paste or summarize the result]

What Worked [Note what the model got right]

What Needs [Note issues: tone, accuracy, structure, hallucination, etc.]

Revision

Final Version [Paste final version after refinement]

Result Rating Strong / Okay / Needs Work

Forward Future: Learn faster. Work smarter. 31

 Tips for Testing Prompts

●​ Change one variable at a time (e.g. role, tone, output format)

●​ Compare multiple outputs with the same prompt using different models
●​ Keep a running prompt library to reuse and adapt over time
●​ If a prompt fails, isolate why: unclear instruction, missing input, poor formatting?

Final Thought

Prompting isn’t about tricks. It’s about clarity. The better you guide the model, the more
valuable it becomes. Once you learn to speak its language, you can do almost anything
— from strategy to storytelling to automation.

This guide is your starting point. Now go make the AI work for you.

Section 7: Glossary of Common Prompt Engineering Terms

Forward Future: Learn faster. Work smarter. 32

This glossary breaks down key terms you've seen throughout the guide — from model
settings like temperature and tokens to advanced concepts like ReAct and Tree of
Thoughts. Whether you're a beginner or brushing up, use this section to stay grounded in
the language of prompting.

Term Definition Example / Context

Prompt The input or instruction you give to a “Summarize this article in 3

language model to generate a response. bullet points.”

LLM (Large Language A type of AI trained on vast amounts of text ChatGPT is an LLM developed
Model) to predict and generate language. Examples by OpenAI.
include ChatGPT, Claude, and Gemini.

Token A chunk of text the model processes, usually “summarize” = 1 token;

a word or word part. Used to count “artificial intelligence” = 2–3
input/output length. tokens

Zero-Shot Prompting Giving the model a task without any “List 3 benefits of this
examples. product.”

Few-Shot Prompting Giving the model examples to help it Example: “Hi, I’m Alex.” →
understand how to respond. “Hello Alex, nice to meet you!”

System Prompt A behind-the-scenes instruction that sets the “Always respond in JSON.”
model’s behavior or output format.

Role Prompting Telling the model to take on a specific “You are a product manager…”
persona or point of view.

Contextual Prompting Including relevant background or data to “Given our brand voice
help the model generate better results. guidelines, write a tweet…”

Step-Back Prompting Asking the model to reflect or solve a “What does the user want?” →
broader question before executing the task. Then write the message

Chain-of-Thought Instructing the model to think through “Think step by step…”

(CoT) reasoning steps before giving a final answer.

Forward Future: Learn faster. Work smarter. 33

 Self-Consistency Running the same prompt multiple times and Used for ambiguous or
selecting the most common or consistent high-stakes tasks.
result.

Tree of Thoughts Exploring multiple reasoning paths or “Explore 3 campaign

(ToT) options before converging on a solution. strategies and evaluate pros
and cons.”

ReAct A prompting framework where the model “Search for X → Summarize →

reasons and then acts — often used with Recommend Y”
tools like web search or code interpreters.

Automatic Prompt Using AI to generate, test, and refine multiple “Write 5 variations of this
Engineering (APE) versions of a prompt. onboarding prompt and rate
them for clarity.”

Temperature A setting that controls randomness. Lower = 0.2 = precise, 0.8 = more
more focused, higher = more creative. varied output

Top-p / Top-k Sampling settings that determine which Useful for tuning creativity vs.
words the model can choose from next — precision.
affects diversity in output.

Max Tokens The maximum length (in tokens) that the Prevents overly long or
model can generate. rambling answers.

Hallucination When the model generates false or made-up “This company raised $50M”
information that sounds correct. — when it actually didn’t.

Format Instruction A prompt that specifies how the output “Respond in JSON” or “Use
should be structured (e.g., list, table, bullet points.”
paragraph).

Persona A defined role or identity assigned to the AI “You are a friendly customer
in a prompt. support agent.”

Forward Future: Learn faster. Work smarter. 34

 Helping everyone benefit from AI with timely ​
news & accessible education.

Website: forwardfuture.ai
YouTube: @matthew_berman
X: @forward_future_

Thanks for reading!​

- Matt, Nick, and the Forward Future Team

Forward Future: Learn faster. Work smarter. 35