liveBook · Manning

Prompt engineering

As explained in Understanding AI

Prompt engineering is the deliberate design and refinement of inputs to guide a model toward desired outputs.

Prompt engineering is the deliberate design and refinement of natural-language instructions that guide a language model toward useful, accurate, and relevant outputs. It is fundamentally a communication skill: translate your intent into a task the model can interpret and execute. section 3.0

Why specificity matters

“Tell me about electric cars” leaves the model guessing whether you want purchase costs, charging infrastructure, maintenance, environmental impact, or driving range. Adding your situation, a specific objective, and delivery instructions reduces that guesswork. section 3.1

Build a well-structured prompt from three complementary parts:

  • Context explains who you are, what you are trying to achieve, your constraints, and what you already know. section 3.1
  • The main point states the specific task or objective you want the model to perform. section 3.1
  • Action controls delivery through format, audience, tone, length, detail, structure, language, or perspective. section 3.1.3
  • Examples can guide behavior through zero-shot, one-shot, or few-shot prompting, while conversations support refinement and exploration. section 3.5
Figure 3.1 The three components of a well-structured prompt, each answering a different question.

Use the three-part prompt structure as a visual checklist: identify the situation, define the task, and specify the response. section 3.1

Table 3.2 Common Prompt Actions
If you want to control…
Examples
Prompt template
Output format
Table, bullet list, report, email, JSON, CSV
Present the answer as [desired format]
Audience
Beginner, high-school student, manager, software engineer, executive
Explain this to [target audience]
Tone
Formal, professional, conversational, persuasive, enthusiastic
Use a [desired tone] tone
Length
One paragraph, 100 words, five bullet points, detailed analysis
Limit the response to [desired length]
Language
English, Italian, Spanish, French
Write the response in [language]
Level of detail
High-level overview, executive summary, technical deep dive
Provide a [desired level of detail]
Structure
Step-by-step instructions, pros and cons, comparison matrix, FAQ
Organize the answer as [desired structure]
Perspective
Cost, sustainability, security, business impact, customer experience
Focus on [specific perspective]

Consult the action patterns when the response needs a particular format, audience, tone, length, language, detail level, structure, or perspective. section 3.1.3

Prompting differs from programming. Programming specifies how a task must be performed; prompting communicates the desired outcome and lets the language model determine how to generate a response from learned patterns. Clear prompts improve results but do not remove hallucinations, overconfidence, bias, prompt injection, or jailbreaking risks. section 3.1.3 section 3.5

How Understanding AI covers it

How deep it goes
Explains prompt structure, response controls, prompting methods, conversation patterns, and limitations, while stopping short of technical model internals.
Examples
Uses an ambiguous electric-car question, prompt-action templates, and the contrast between prompting and an explicit battery-warning program.
What it assumes
Assumes natural-language interaction but teaches prompt structure and model communication from the basic idea of an instruction.
What you can do after
Apply it via language models by specifying context, objectives, response actions, examples, and verification needs.
How it connects
Connects context, intent, language models, zero-shot prompting, few-shot prompting, prompt injection, jailbreaking, and human oversight.

Also explained in

Compare how 51 books explain prompt engineering →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Understanding AI. The section links point to the parts of the book it draws on.