A few weeks ago I came across an infographic comparing 8 different prompting frameworks for AI. I'd heard of most of them, used a couple regularly, and completely ignored the rest. So I did what any reasonable person would do: I sat down and actually tested all of them on real tasks. Here's what I found.

Spoiler: there's no single winner. But there are clear patterns for when each one works well and when it just wastes your time. Let's go through them one by one.

1. RISEN

Role, Instructions, Steps, End goal, Narrowing. This is one of the more structured frameworks and it shows. RISEN forces you to think about context before you type a single word of the actual request.

When to use it: Complex tasks where the AI keeps missing the point. Anything that involves multiple steps or requires specific expertise — like writing a technical audit, drafting a legal document, or generating a detailed plan.

Example: "Role: You are a senior UX designer with 10 years of SaaS experience. Instructions: Review the following landing page copy. Steps: 1) Identify clarity issues, 2) Flag weak CTAs, 3) Suggest rewrites. End goal: A landing page that converts better. Narrowing: Focus only on copy, not visual design."

I use RISEN when I need something done properly the first time. It's a bit verbose to write, but it cuts down on back-and-forth dramatically.

2. Chain-of-Thought (CoT)

"Let's think step by step." Sounds almost too simple, right? But it works. Chain-of-Thought prompting encourages the model to reason through a problem before jumping to an answer. This reduces the classic AI problem of confidently stating something wrong.

When to use it: Math, logic puzzles, troubleshooting, any multi-step reasoning. Also surprisingly useful for ethical dilemmas and "what should I do" business decisions.

Example: "I need to decide whether to hire a full-time developer or use a freelancer for this project. Think step by step through the trade-offs considering: project timeline (3 months), budget (20k PLN), and our need for ongoing support."

CoT is in my daily toolkit. It's the first thing I try when I need the AI to think rather than just generate.

3. Few-Shot

Show, don't tell. Instead of explaining what you want, you give 2-3 examples and let the model figure out the pattern. This is especially powerful for consistent formatting, tone matching, or repeating tasks.

When to use it: Anything where you need consistent output — social media posts, product descriptions, customer emails. Also great for teaching the AI your personal writing style.

Example: "Here are three examples of my blog intro style:
[Example 1]
[Example 2]
[Example 3]
Now write an intro for an article about Google Business Profile optimization."

I use Few-Shot constantly for client content. Once you've got good examples saved, you can regenerate consistent copy in seconds.

4. PROSTO

Persona, Role, Objective, Scope, Task, Output. This is a Polish-made framework (the name means "simple/straight" in Polish) created specifically for business use cases. I've been testing it heavily over the past few months and it's become one of my favorites.

When to use it: Business communications, marketing copy, strategic planning. PROSTO works well when you need the AI to think from a specific business perspective and produce structured, client-ready output.

Example: "Persona: Small business owner, 10 employees, service sector. Role: You are a business advisor specializing in digital marketing. Objective: Help me increase local visibility. Scope: Focus on free and low-cost tactics. Task: Create a 30-day action plan. Output: Bulleted checklist, grouped by week."

What I like about PROSTO is how it separates persona (who you're writing for) from role (who the AI should be). That distinction matters more than you'd think.

5. RTF — Role, Task, Format

The minimalist's framework. Three words, three fields. RTF is fast to write and surprisingly effective for straightforward tasks.

When to use it: Quick requests where you know exactly what you want. Best for simple content generation, summaries, translations.

Example: "Role: Marketing copywriter. Task: Rewrite this product description to be more conversational. Format: 3 short paragraphs, max 100 words each."

RTF is my go-to for quick daily tasks. When RISEN feels like overkill, RTF gets the job done.

6. TAG — Task, Action, Goal

Similar to RTF but with a stronger emphasis on the end goal. TAG keeps you honest about what you actually want to achieve, not just what you want the AI to do.

When to use it: When you keep getting technically correct but strategically wrong answers. TAG forces you to state the "why" behind the task.

Example: "Task: Analyze our website homepage. Action: List 5 specific improvements. Goal: Increase the number of people who contact us through the contact form."

Good framework for conversion-focused work. The Goal field changes everything — the AI gives completely different advice when it knows the business objective.

7. RACE — Role, Action, Context, Execute

RACE adds Context as a first-class element, which makes it useful when background information is critical to getting the right answer.

When to use it: Industry-specific tasks, situations where the AI needs to understand your unique circumstances before making recommendations.

Example: "Role: Local SEO specialist. Action: Create a Google Business Profile optimization checklist. Context: The business is a plumbing company in a mid-size Polish city, competing with 5 established local plumbers, with no existing online reviews. Execute: Output as a prioritized checklist with estimated time per item."

RACE shines when context is everything. A generic answer without context is often useless — this framework prevents that.

8. BAB — Before, After, Bridge

A copywriting classic adapted for AI. BAB is built around transformation: where you are, where you want to be, and what gets you there.

When to use it: Marketing copy, sales emails, landing page content. Anything where you need to create emotional resonance around a product or service.

Example: "Before: Our client spends 3 hours a week manually posting to social media and gets almost no engagement. After: With our tool, they schedule a week of content in 20 minutes and reach 3x more people. Bridge: Write a Facebook ad that connects these two states and leads to a free trial signup."

BAB is the most powerful framework for persuasive writing. If you're writing anything that needs to move someone to action, start here.

Which ones do I actually use?

Honestly? About three of these are in my daily rotation.

Chain-of-Thought for anything that requires reasoning. I literally add "think step by step" to prompts all the time now. It's a reflex.

Few-Shot for content that needs to match a specific style or format. Client emails, social posts, product descriptions — always with examples.

RISEN for complex, high-stakes tasks where I need to get it right the first time. It takes longer to write the prompt but saves hours downstream.

The others I use situationally. PROSTO has been growing on me for business strategy tasks. BAB is my default for sales copy. RTF handles everything else.

The real takeaway

These frameworks aren't magic. They're structured ways of giving the AI the information it needs to give you a useful answer. The more context and constraints you provide, the better the output.

You don't need to pick one and stick with it. Learn 2-3, understand what problem each one solves, and mix them as needed. The goal is better results, not framework purity.

And if none of them work? Try just having a conversation. Sometimes the best prompt is just talking to the AI like you'd talk to a smart colleague.

Originally published in Polish: 8 frameworków promptowania AI — który wybrać?