Mandragore Learning Academy
Careers & jobs

Prompt Engineering: methods, examples and professional skills

Prompt Engineering is the practice of designing and improving instructions for AI models so outputs are more relevant, reliable and easier to verify. It focuses on task clarity, context, constraints, output format and evaluation.

Who is this article for?

For professionals and learners who want to understand this skill, its practical uses and the progression required before choosing or continuing training.

What you need to understand

Prompt Engineering is the practice of designing and improving instructions for AI models so outputs are more relevant, reliable and easier to verify. It focuses on task clarity, context, constraints, output format and evaluation.

The skill must be connected to a business need, identified data or information, an expected outcome and control criteria. Professional mastery is measured by reliable, explainable results rather than the number of features known.

Why this skill matters today

This skill supports growing needs around productivity, data quality, analysis, automation and changing job practices. Its value still depends on the use case and on integration into a real process.

For organisations, the priority is therefore to build method, controls and real-world practice rather than simple tool familiarity.

What you actually do with this skill

Clarify tasksTurn a vague request into an objective, audience, constraints, success criteria and clearly stated out-of-scope items.
Structure contextProvide only the useful context: data, business rules, definitions, examples and limits required to complete the task.
Specify formatsRequest an actionable structure—JSON, table, plan or mandatory fields—so the output can be checked and reused.
Use examplesProvide one or two representative examples to clarify expected quality, format or decision logic.
Evaluate and iterateCompare the output with explicit criteria, identify defects and improve the prompt or context instead of retrying blindly.

Putting the skill into practice

A professional needs to solve a recurring task with an outcome that can be checked and handed over.

Problem : Reduce processing time without losing quality.

Context : Real data, business constraints and validation rules.

Skills : Objective, Context, Constraints, Format, Few-shot.

Example : Clarify tasks inside a defined workflow.

Hands-on application : Test several cases, document limitations and keep the appropriate validation step.

Skills to master

The skill must be connected to a business need, identified data or information, an expected outcome and control criteria. Professional mastery is measured by reliable, explainable results rather than the number of features known.

BeginnerUnderstand foundations, vocabulary and the core workflow.
IntermediateApply the skill to realistic cases and combine related tools.
Advanced / autonomousDesign, control, document and improve the process autonomously.

Errors, limitations and points of attention

  • Starting with the tool before the problem.
  • Skipping verification of important outputs.
  • Ignoring data quality or access rights.
  • Building a process that cannot be reproduced.
  • Failing to document assumptions and limitations.

How to progress

BeginnerUnderstand foundations, vocabulary and the core workflow.
IntermediateApply the skill to realistic cases and combine related tools.
Advanced / autonomousDesign, control, document and improve the process autonomously.
Mandragore Learning Academy

Train with Mandragore Learning Academy

Mandragore Learning Academy connects this topic to training that actually exists in its catalogue. Courses are structured to progress from foundations to controlled professional practice.

The logic remains the same: understand, see a real case, know what to learn, then practise in the Learning Workspace.

Related training path

Related training path

Mandragore Learning Academy connects this topic to training that actually exists in its catalogue. Courses are structured to progress from foundations to controlled professional practice.

Sources and references

Mandragore source: config_ia.php, config_dev.php and config_microsoft.php remain the source of truth for the training actually offered.

  • OpenAI — Prompt engineering best practices for ChatGPT
  • Anthropic — Prompting best practices