Prompt Engineering 101

Introduction

‍

Crafting better prompts = building better agents

This chapter gives you the foundation to master the art of writing prompts and unlock your AI agents’ full potential , no coding needed.

1. What is prompt engineering?

‍

Prompt engineering is the art of designing and refining instructions to guide large language models (LLMs) effectively. For product managers, it’s a superpower: it allows you to prototype, test, and iterate on AI use cases without writing a single line of code.

When done right, prompt engineering helps translate product intent into actionable AI behaviors. When done wrong, it leads to hallucinations, inconsistencies, and degraded UX.

Why PMs should care:

1. It reduces dependency on engineers for testing or prototyping.
2. It improves the output quality of AI-powered features.
3. It bridges the gap between business needs and LLM capabilities.

Good prompts are clear, contextual, and structured. Here’s how to make your prompts reliable and reproducible.

Prompt Library — Basalt 🗿

Clarity : Avoid ambiguity. Use explicit language. Specify the task and expected output.

✅ “ Summarize this customer call in 3 bullet points: their problem, emotions, and the proposed solution. ”

Context : LLMs don’t “know” your world unless you tell them. Set the scene:

✅ “ You are a product analyst. Summarize this user research session for the leadership team. ”

Format specificity : Guide the structure of the output.

✅ “ Return a table of the 5 most requested features with user count and priority (1–5). ” provide exemples for best results !

Iteration : There’s no perfect prompt on the first try. Test, compare, refine. Prompt engineering is an iterative loop.

3. Essential techniques to master

‍

Prompting is a skillset — not a hack. Here are a few repeatable techniques that work for most PM use cases.

Mastering these practical prompt techniques will elevate your AI interactions from basic queries to precise, reliable, and context-aware responses — essential skills for every PM building AI agents.

• 💉  Few-shot prompting  : donne au modèle un court « tutoriel » avec 2 à 5 exemples d’entrée/sortie pour l’aider à comprendre la structure et le style attendus, ce qui améliore la précision par rapport au zero-shot.
• 🧠  Chain of Thought (CoT)  : encourage le modèle à détailler son raisonnement étape par étape avant de répondre, pour plus de fiabilité et d’exactitude.
• 🎭  Role prompting  : assigne un rôle précis au modèle pour améliorer l’alignement du contexte et la pertinence des réponses.
• 📋  Format specification  : indique clairement le format ou la structure attendue dans la réponse pour réduire ambiguïtés et erreurs.
• 📝  Self-critique & evaluation  : demande au modèle de réfléchir, critiquer ou justifier sa propre réponse pour un meilleur contrôle qualité.
• ⛔  Negative instructions  : indique explicitement ce qu’il faut éviter pour prévenir un contenu hors sujet ou indésirable.

Focus on : Prompt chaining

Prompt chaining concept

It enables you to:

• Divide complex tasks into manageable subtasks.
• Improve response coherence and customization.
• Make the AI’s reasoning clearer and easier to control.

For example, to generate a report, you might first ask the AI to identify key points, then summarize each one, and finally compile everything into a structured document. Each step uses a separate, linked prompt.

Prompt chaining is especially useful for multi-step tasks like writing, data analysis, or problem solving, and helps build more reliable AI assistants.

4. Why this matters: the strategic value for PMs

‍

Prompt engineering is not just tactical — it’s strategic. It lets PMs lead AI initiatives without waiting on technical teams.

Prompt engineering goes far beyond being a mere technical skill — it is a strategic lever that empowers product managers to take ownership of AI initiatives and accelerate product innovation without being bottlenecked by engineering teams.

By mastering prompt design, PMs gain the ability to rapidly prototype AI features, translate complex business needs into precise AI instructions, and significantly improve the consistency and reliability of AI outputs. This skill positions PMs not just as facilitators, but as true leaders in shaping AI-driven product experiences.

Key benefits include:

1. Accelerated prototyping and validation :  Instead of waiting weeks for engineering cycles, PMs can quickly turn ideas into working AI demos, enabling fast experimentation and early user feedback. This “build-first, code-later” approach drastically shortens innovation cycles.
2. Enhanced cross-functional communication :  Prompt engineering acts as a universal language that bridges the gap between messy, ambiguous business goals and the precise, technical instructions AI models require. This clarity streamlines collaboration across product, design, and engineering teams.
3. Improved model reliability and user trust :  Carefully crafted prompts reduce hallucinations, errors, and unpredictable behaviors from AI models. Clear instructions lead to outputs that align better with user expectations and business rules, boosting confidence internally and externally.
4. Ownership of AI product vision :  PMs skilled in prompting move from passive stakeholders to active creators who directly shape the AI behavior embedded in their products. They can define how AI delivers value, ensuring alignment with customer needs and company strategy.