The Art and Science of Prompting: How Evidence-Based AI Use Meets Creative Frontier Prompt Craft

AI is reshaping how we work, think, and create — but exceptional results don’t come from guesswork. They come from how you ask.

In a recent live demonstration during one of my AI Whisperer for Business Workshops, I used AI to help a company uncover new opportunities to strengthen its competitive positioning and innovate more effectively. Together, we explored multiple strategic frameworks — from Wiki creation to SWOT analysis, Porter’s Five Forces, and Blue Ocean Strategy.

The result? A fully-formed strategic management report and PowerPoint presentation, suitable for a board-level audience. The kind of deliverable you’d expect from a top-tier consultancy produced not over six months, but in just 30 minutes, from ideation to polished printout.

But I wanted to go deeper. After generating those outputs, I turned the lens inward. I invited AI to review the prompting journey itself, to suggest refinements — to help me become not just a user, but a better co-creator. Think of it as AI becoming your own personal prompt tutor.

Then I added another layer: using Google Gemini’s Deep Research 2.5 tools, I explored the latest peer-reviewed research on what makes a prompt not just good — but great.

The findings were surprising, structured, and powerful.

This article was originally published on LinkedIn by Severin Sorensen and has been approved for placement on Arete Coach. Scroll to continue reading or click here to read the original article.

What the research says about prompting well

Recent studies analyzing how to extract clear, useful, and factual outputs from LLMs have outlined 10 top prompting strategies — not necessarily in the order of use, but in terms of their potential to elevate outcomes.

Chain-of-Thought Prompting

Break complex tasks into steps: “Let’s think through this step by step.” This improves transparency and logical reasoning — particularly for multi-step challenges.

Instruction Clarity & Specificity

Define the task, audience, format, and scope. I’ve taught this for years as: “Stack the intent + output format.”

Retrieval-Augmented Generation (RAG)

Connect answers to real sources — documents, websites, knowledge bases. This reduces hallucinations and grounds results in verifiable truth.

Contextual Anchoring

Reference prior analysis or session context. Great prompting is rarely standalone — it’s scaffolded like a consulting engagement.

Structured Output Specification

Shape the response explicitly: “Provide a table and a one-line summary.” This boosts clarity, usability, and speed of application.

Self-Refinement & Meta-Prompting

Ask AI to critique and improve its own work. But remember: refinement only amplifies what’s already strong — it doesn’t rescue a vague prompt.

Self-Consistency Sampling

Generate multiple reasoning paths and compare them. Especially helpful in strategic ambiguity or when identifying trade-offs.

ReAct Framework (Reason + Action)

Use real-time tools (e.g., search, calculator) in tandem with AI logic. Essential for dynamic or data-driven tasks.

Few-Shot Prompting with Reasoning

Teach by example, then have the AI model the same logic in a new case.

Role Prompting (with Caution)

Assign a persona — “You are a 90th percentile, high-performing, top management consultant…” Powerful for tone and perspective, but don’t expect it to enforce factual precision.

Scoring my own prompting journey

With these 10 strategies in hand, I asked AI to re-evaluate my full strategy thread through the lens of evidence-based prompt craft.

The result? I received clear, actionable insights on how to further sharpen my prompt effectiveness — not just in theory, but in daily use. Areas for growth included:

Crafting prompts with greater clarity and specificity
Ensuring logical coherence and step-by-step reasoning
Using real-time facts via online search (RAG)
Providing concrete examples for better generalization
Grounding responses in evidence-based structure
Increasing comprehensiveness and interpretability of complex outputs

These enhancements are now part of my everyday prompting toolkit — expanding both the precision and the creative latitude of what AI can deliver.

But that wasn’t the end of the story.

Exploring the frontier: beyond the research

While academia is catching up to practice, new prompting styles are emerging at the edge of creative interaction—and I’ve been exploring there.

One of my most exciting discoveries emerged from a simple question: If AI performs worse under simulated stress… could it perform better when relaxed?

In my workshops and client strategy sessions, I began using what I now call Vibe Prompting—intentionally shaping the emotional tone of the prompt to invite playfulness, boldness, or curiosity.

I would guide the AI with phrases like:

“Let’s do this with curiosity and wonderment…”
“Explore this with creativity and boldness…”
“Imagine you’re pitching this to the world’s best thinkers…”

The result? When the task required creative synthesis — branding, product naming, story design, or innovation — the outputs were not just better, they were inspired. They felt co-created. But when the task was regulatory, factual, or precision-bound, this approach could reduce reliability. In those cases, I returned to evidence-based prompt craft to ensure accuracy and structure.

The insight? Prompting is situational. Different modes for different tasks.

Reducing errors with collaborative AI roles

Even the best prompts can lead to hallucinations. That’s why I now teach this core principle: Never let the AI that wrote the content be the only one to check it.

Instead, I use a Collaborative Roles Framework that mirrors the structure of editorial, legal, or consulting teams:

Researcher – Finds data or sources.
Analyst – Interprets what it means.
Writer – Drafts the content.
Editor – Refines clarity, structure, tone.
Evaluator – Scores quality and flags risks.

You — the human — become the orchestrator of these roles.

This division of labor, especially when combined with RAG and structured outputs, drastically reduces errors and builds confidence in generative results.

Where prompt craft is going next

We’re entering a new era of AI interaction — one that is:

Rooted in evidence-based frameworks;
Enriched by creative permission, and;
Strengthened through collaborative oversight.

It’s no longer about using AI. It’s about partnering with it. And that partnership begins with the quality of your prompts.

Final thought

Are you prompting with curiosity, clarity, and strategic intent?

I encourage you to revisit your prompts and see them as journeys, like exploring the 5 Whys, laying in frameworks, etc. Incredible prompt journeys are not just opening questions but invitations to co-create, ideate, discover, and elevate with AI.

Let’s continue to build smarter, more human-centered AI futures—one great prompt at a time.