RADICAL Prompting
The framework for AI results that make the difference — 7 steps from vague request to expert prompt.
RADICAL is a 7-step framework for effective AI prompting: Role, Ask, Data & Context, Industry, Constraints, Answer Format, Level. Developed by Radical Innovators, it works on ChatGPT, Claude, Gemini, and any other AI tool. This article demonstrates the framework live with a hospitality example and shows why structured prompting is the only AI skill that never becomes obsolete.
75% use AI. 36% have ever been trained.
Same AI, same access, same subscription — and yet the results are worlds apart. One prompt delivers a generic paragraph nobody can use. The next produces a board-ready strategy. The difference is not the tool. The difference is the question.
What research shows
in losses threaten the global economy from the AI skills gap by 2026. Over 90% of companies face a critical skills gap according to IDC. Only one third of employees received AI training in the past year — even though 78% of companies already use AI. The gap between AI adoption and AI competency is growing faster than any other technology skills gap.
Why the same AI delivers brilliance for some — and mediocrity for others
An executive types "Write me a strategy for Q3" into ChatGPT and gets three paragraphs of platitudes. Their colleague frames the same task with a role, context, industry data, and clear constraints — and receives a document that goes straight into the board meeting. No talent gap. No better tool. Just structure.
The tool landscape shifts every three months: ChatGPT, Claude, Gemini, Copilot, Llama, Cursor, Claude Code — anyone betting on a single product starts over every few months. Anyone who can prompt structurally transfers that skill to every tool. That is the one AI competency that never becomes obsolete.
What research shows
higher quality was delivered by consultants with AI support — and they were 25% faster. But only when they interacted effectively with the model. The study examined 758 consultants on real consulting tasks and showed: without structured prompts, results remained below the level of experienced consultants without AI. These findings were confirmed in 2025 by McKinsey's "State of AI" study: 78% of companies use AI, but only 1% describe themselves as truly mature. The difference between the top performers (6% of companies) and the rest: structured processes — not better tools.
RADICAL: 7 Steps to the Expert Prompt
Seven letters. Seven steps. One system that works on every large language model — from ChatGPT to Claude to Gemini. The acronym is deliberate: R-A-D-I-C-A-L. Defining industry knowledge as its own explicit step rather than letting it get lost in general context — that's what separates RADICAL from generic frameworks like CRAFT or COSTAR.
Role.
Define who the AI should be. Not "you are an assistant," but "you are a revenue manager at a 4-star hotel with 10 years of experience in the European market." The more precise the role, the more relevant the output.
Ask.
Name the exact deliverable. Not "help me with my strategy," but "create a competitive analysis with 3 scenarios and concrete action items." Vague requests produce vague results.
Data & Context.
Feed the AI everything relevant: numbers, history, market data, guest profiles, past results. AI cannot guess what you know — it needs your data to solve your problems.
Industry.
The step that makes the difference. Industry-specific knowledge changes everything: regulation (GDPR, NIS2, PCI-DSS), domain language, market dynamics, competitive landscape. A prompt without industry context delivers generic answers.
Constraints.
Set clear boundaries: character limit, tone of voice, compliance requirements, what the AI should NOT do. Without constraints, every AI produces endless, unusable output.
Answer Format.
Define the structure: table, email template, checklist, bullet points, prose. The format determines usability — perfect content in the wrong format is worthless.
Level.
Set the quality benchmark: executive summary vs. operational playbook vs. print-ready document. Tell the AI who the output is for and at what level it should be delivered.
Not every task needs all seven steps. For a quick email, R + A will do. For a competitive analysis, you need R + A + D + I. For a board-ready document, all seven. Rule of thumb: the higher the consequence of a bad answer, the more steps are worth it.
Other frameworks like CRAFT or COSTAR cover role, task, and context — but industry knowledge gets lost in the general context. The I step (Industry) as its own explicit building block is the reason a RADICAL prompt for a Vienna luxury hotel sounds different from one for a mechanical engineering company in Munich — and why both deliver results that a generic framework cannot reach.
RADICAL in 60 Seconds
One prompt. Seven building blocks. Here's a hotel review response — built in under a minute.
Respond to this negative Google review professionally. The guest complains about construction noise and cold breakfast but praises the friendly reception.
R: Guest Relations Manager, 4-star hotel Vienna · A: Response to 2-star review (noise, cold breakfast, friendly reception) · D: Construction until June 2026, courtyard upgrade, food warmer replaced, earplugs and upgrade to quiet courtyard rooms as compensation · I: Luxury segment Vienna · C: Max 180 words, formal, no discounts · A: Ready-to-post Google response · L: Individual response
Dear Guest, thank you for your feedback. We are sorry that your stay did not meet your expectations. We take your criticism very seriously and will discuss the issues internally. We hope to welcome you back soon. Kind regards, Your Hotel Team
Dear Guest, thank you for your candid review — and for your kind words about our reception team. The construction next door is a construction project running until June 2026. We offer affected guests earplugs and a complimentary upgrade to our quiet courtyard rooms. The breakfast warmer has already been replaced. Please contact us directly — we would like to make it right.
Same effort, same tool. But one response names concrete solutions instead of hiding behind platitudes.
Same effort. Same tool. But a result that reads like the hotel director wrote it personally.
Five Industries. Five Results.
RADICAL works everywhere. Not every task needs all seven steps — for simple requests, R + A will do. The higher the stakes, the more steps pay off.
Financial Analysis — Board Reporting
Create a quarterly analysis for Q1. Revenue 47.2M, EBIT margin 12.3% vs. planned 14.1%. What are the main drivers?
R: CFO analyst, engineering company 280 employees · A: Q1 analysis with variance commentary · D: €47.2M (+8% YoY), EBIT 12.3% (Plan: 14.1%), raw materials +22% · I: Engineering DACH, IFRS, supervisory board · C: Max 2 pages, no speculation · A: Board presentation · L: Supervisory board level
The Q1 analysis shows revenue of €47.2M. The EBIT margin is 12.3%, below the planned 14.1%. This indicates increased costs. Recommendation: The cost structure should be analyzed and optimization potential identified.
Executive Summary: Q1 2026 revenue at €47.2M (+8% YoY). EBIT margin 12.3% — 180 basis points below plan. Primary driver: raw material costs (+22%). Proposed actions: (1) Review price escalation clauses in existing contracts, (2) Evaluate commodity hedging for Q2/Q3, (3) Cost structure review with procurement and production. Detailed analysis by business unit on page 2.
The generic output parrots back your own numbers. The RADICAL output delivers a board-ready brief with concrete action items.
Banking — Compliance Documentation
Create a summary of the new DORA requirements for our IT department.
R: Regulatory analyst, Sparkasse, 1,200 employees · A: DORA compliance gap analysis for IT department · D: Current status: ISO 27001 certified, no dedicated ICT risk management, 340 IT systems in use · I: German banking sector, BaFin supervision, DORA Art. 5-16 (ICT risk management), Art. 17-23 (incident reporting with 4h initial notification deadline), DORA deadline January 2025, EBA guidelines · C: Max. 3 pages, no legal advice, traffic-light format (Green/Yellow/Red) for gap status · A: Board presentation with action plan · L: Audit-proof
DORA stands for Digital Operational Resilience Act and takes effect in January 2025. Banks must make their IT systems more resilient. The key requirements concern ICT risk management, incident reporting, and third-party risk management. It is advisable to begin implementation early.
Gap analysis DORA compliance: ICT risk management (RED — no dedicated framework, DORA Art. 5-16 requires formal ICT risk framework by Q1 2025). Incident reporting (YELLOW — process exists but reporting deadlines not DORA-compliant: 4h initial notification missing). Third-party risk (YELLOW — 340 IT systems without documented ICT third-party management). Action plan: 3 priority-1 measures with timeline on page 2.
Generic: Wikipedia summary with no relevance to the actual situation. RADICAL: A gap analysis with traffic-light status that the board understands immediately.
Hospitality — Seasonal Campaign
Create 3 summer packages for our lake resort in different price tiers for families and couples.
R: Revenue marketing strategist · A: 3 packages (budget/standard/premium) · D: 4-star, 75 rooms, ADR €175, spa treatments, e-bike rental, half board available, childcare available · I: Carinthian lakes, competition from holiday rentals · C: Min 15% above standard rate, bookable online · A: Name, target group, inclusions, price, headline · L: Business case level
Package 1 "Summer Dream": Overnight stay incl. breakfast and access to the wellness area. Package 2 "Family Fun": Half board and childcare. Package 3 "Premium": Suite with lake view and dinner.
Premium "Lake View": €289/night, suite, dinner, 2x spa, e-bikes, late checkout. Target: couples. Standard "Family Time": €219/night, family room, half board, childcare. Budget "Summer Fresh": €209/night, double room, breakfast, lake access. All packages bookable online.
Without RADICAL: three package names with no prices, no target audience, no business case. With RADICAL: a decision-ready brief for management.
E-Commerce — Product Descriptions
Write SEO-optimized product descriptions for our premium outdoor jackets made of Gore-Tex Pro, about 150 words.
R: E-commerce copywriter, SEO · A: Product description outdoor jacket · D: Gore-Tex Pro 3-layer, 28,000mm water column, 450g, YKK Aquaguard, helmet-compatible hood, Napoleon pocket, made in Portugal, price €429 · I: D2C outdoor, competition Arc'teryx/Mammut · C: 150 words, keywords natural · A: Shopify + meta title/description · L: Conversion-optimized
This high-quality outdoor jacket made of Gore-Tex Pro offers optimal protection in any weather. The innovative material is waterproof yet breathable. The jacket is perfect for hiking, climbing, and skiing. Order now and enjoy the outdoors!
Alpine Pro 3L — For conditions where others turn back. 28,000mm water column, 450g, made in Portugal. YKK Aquaguard zippers, helmet-compatible hood, Napoleon pocket. Not the lightest on the market. But the one you won't regret on a ridge in driving rain. From €429.
Generic: assembly-line copy that fits any jacket. RADICAL: positioned against Arc'teryx with real specs and a reason to pay €429.
Social Media — Instagram Content
Write an Instagram post for our new sofa. Premium furniture brand, walnut, handcrafted.
R: Social media strategist, D2C premium furniture · A: Instagram carousel post (5 slides) for walnut sofa "Lina" · D: Price €3,890, handcrafted in Austria, FSC walnut, 12-week lead time, 89 reviews (4.8 stars), approx. 250 hours production time per piece, solid wood (no veneer) · I: D2C premium furniture, competition Vitra/USM, target audience design-conscious 30-50 · C: Carousel slides, no "buy now", storytelling about craftsmanship, max 150 words caption · A: 5 slide descriptions + caption + 10 hashtags · L: Brand level, no performance marketing
Our new walnut sofa combines timeless design with supreme comfort. Handcrafted with attention to detail. Discover our premium collection now! #furniture #design #premium #sofa #interior
Slide 1: 250 hours of handcraft. One sofa. (Close-up of walnut grain) Slide 2: FSC walnut from Austrian forestry. No veneer. Solid. Slide 3: "Lina" — designed for rooms you never want to leave. Slide 4: 89 reviews, 4.8 stars. Slide 5: €3,890. 12 weeks. Handcrafted in Austria. Caption: Some furniture you buy. Some you have built. #premiumfurniture #handcrafted #walnut #austria #designsofa #interiordesign #solidwood #livingroom #furnituredesign #madeinaustria
Generic: advertising clichés that fit any piece of furniture. RADICAL: Storytelling with concrete numbers that stands out from Vitra and USM.
Five Mistakes That RADICAL Prevents
No role → generic output.
Without defined expertise, the AI responds as a generalist. "You are Revenue Manager" changes everything.
Vague task → irrelevant result.
"Help me with marketing" produces textbook answers. A concrete deliverable forces substance.
No context → hallucinated facts.
Without your data, the AI invents plausible but false numbers. Your data in, your results out.
Industry ignored → wrong terminology.
A prompt without industry context confuses hospitality KPIs with retail metrics.
No constraints → endless output.
Without boundaries, the AI delivers 2,000 words when you need 200.
What research shows
accuracy on mathematical tasks rises through structured prompting alone (Self-Consistency, Wang et al. 2022). Chain-of-Thought prompting improves arithmetic accuracy by 58% (Wei et al., NeurIPS 2022). The quality of the prompt is the single largest lever for output quality — greater than the choice of model.
GDPR, EU AI Act, and Self-Hosted AI
The elephant in the room: Can we actually enter company data into ChatGPT? The answer is nuanced — and the D step in RADICAL (Data & Context) becomes critical here. Anyone prompting with sensitive customer data, financial figures, or health information needs more than good intentions.
The solution: Self-hosted or hybrid AI models. Open-source models like Llama, Mistral, or Mixtral run on your own infrastructure — no data leaves the company. Cloud providers like Azure OpenAI Service or AWS Bedrock offer enterprise options with data processing agreements and EU data centers. For many European companies, this is the only GDPR-compliant path.
Practical recommendation: Define three data categories for AI usage. Green: public information, market data, general text — any tool works. Yellow: internal processes, aggregated KPIs — enterprise versions with DPA only. Red: personal data, trade secrets, health data — self-hosted models only or no AI at all.
The EU AI Act (taking effect in stages since February 2025) classifies AI systems by risk level. For high-risk areas like HR screening or credit decisions, strict documentation and transparency requirements apply. RADICAL helps here too: the C step (Constraints) is where compliance requirements flow directly into the prompt — "Do not use personal data," "Justify the decision," "Run bias check."
The good news: Structured prompting and data protection are not contradictory. Quite the opposite — anyone who takes the D step seriously automatically thinks about which data belongs in the prompt and which does not. That is data protection by design.
What research shows
in GDPR fines have been issued since 2018 — in 2025 alone, that figure was €1.2 billion. 443 data breaches are reported daily (a 22% increase year-on-year). The EU AI Act raises the stakes further: up to €35 million or 7% of global annual revenue for prohibited AI practices. For European companies, this means: data sovereignty is not a nice-to-have — it is a compliance obligation.
The RADICAL framework as a printable quick reference — all 7 steps, prompt template, power tips and data privacy guide on a single page.
How We Bring RADICAL to Organizations
Radical Innovators implements the RADICAL framework through workshops, team trainings, and industry-specific prompt libraries. No PowerPoint lectures — hands-on with real company data, real tasks, real results. From half-day executive briefings to multi-week team programs. Every industry gets its own I step: with the relevant regulations, the right domain language, and the metrics that matter in that sector.
The question is no longer whether your team uses AI — but how well. RADICAL gives you the structure to achieve better results on any tool, in any industry, for any task. This is not theory. This is a tool.
— Martin Kocijaz, CEO Radical Innovators