CRISPE Framework
Six dimensions, one prompt. CRISPE gives the AI a complete briefing — who to be, what to know, what to do, how to sound, and how many ways to try it — so every response arrives fully formed and ready to use.
Introduced: CRISPE emerged in 2023 as a community-developed structured prompting framework. It evolved from the growing practitioner consensus that effective AI prompts need more than just a task description — they need to define the AI’s capacity, supply relevant background, specify the request, set a communication style, and encourage exploration of multiple approaches. The acronym captures six dimensions: Capacity and Role, Insight, Statement, Personality, and Experiment.
Modern LLM Status: CRISPE remains highly relevant and practical for modern LLMs. While Claude, GPT-4, and Gemini can produce useful responses from minimal prompts, they consistently deliver superior results when given the structured context CRISPE provides. The framework’s “Experiment” dimension is particularly valuable today — requesting multiple variations from a single prompt leverages the probabilistic nature of language models and gives users genuine options rather than a single take-it-or-leave-it answer.
Brief the AI Like You’d Brief a Consultant
When you hire an expert consultant, you don’t just hand them a task and walk away. You tell them what role to play, share the background they need, describe exactly what you want, explain the tone and style that fits your audience, and ask them to present a few different approaches so you can choose. CRISPE applies this same professional briefing structure to AI prompts.
Each letter addresses a different dimension of the request. Capacity and Role define who the AI should be. Insight provides the knowledge it needs. Statement specifies the task. Personality shapes how the response should sound. Experiment asks for multiple variations — turning a single prompt into a creative exploration rather than a one-shot guess.
The result is a prompt that leaves almost nothing to chance. Instead of hoping the AI interprets your intent correctly, you’ve spelled out every dimension that matters — from expertise level to communication style to the number of alternatives you want to evaluate.
A simple prompt like “Write me a marketing email” forces the AI to guess at expertise level, tone, context, and format. CRISPE eliminates these guesses by addressing each dimension explicitly. The Experiment component is especially powerful: instead of accepting the first output, you ask the AI to generate multiple approaches — giving you options to compare, combine, or refine. This turns prompt engineering from a single roll of the dice into a structured exploration.
The CRISPE Process
Six steps from blank prompt to comprehensive AI briefing
C — Capacity and Role
Define what role the AI should assume and what capabilities it should bring to the task. This goes beyond simple role assignment — you specify the depth of expertise, the professional context, and the scope of authority the AI should operate within. A well-defined capacity sets the foundation for everything that follows.
“Act as a senior data scientist with 15 years of experience in predictive analytics for e-commerce, specializing in customer churn modeling and lifetime value prediction.”
R — (Part of Capacity)
In CRISPE, the “R” is embedded within the Capacity component rather than standing alone. The role is inseparable from capacity — what the AI can do and who it should be are defined together. This unified approach prevents contradictions between assigned expertise and assigned persona that can occur when role and capability are specified separately.
“You are the lead analytics consultant presenting findings to a non-technical executive team — translate complex statistical concepts into business impact language.”
I — Insight
Provide the background information, context, and domain knowledge the AI needs to produce an informed response. Insight bridges the gap between the AI’s general training and your specific situation. The more relevant context you supply, the less the AI has to guess — and guessing is where hallucination and irrelevance creep in.
“Our SaaS platform has 50,000 active users. Monthly churn is 4.2%, up from 3.1% last quarter. Our primary competitors recently launched free tiers. Customer support tickets have increased 30% in the last 60 days, mostly about our pricing structure.”
S — Statement
State the specific task, question, or deliverable you want. The Statement is the core of your request — what you actually need the AI to produce. Keep it clear and actionable. Everything else in CRISPE exists to support this central ask, ensuring the AI has the context, persona, and style guidance to deliver exactly what you need.
“Develop a 90-day churn reduction strategy with specific tactical recommendations, projected impact metrics, and a prioritized implementation roadmap.”
P — Personality
Define the tone, style, and communication approach for the response. Personality controls how the AI delivers its answer — whether formal or conversational, technical or accessible, concise or comprehensive. This dimension ensures the output is not just correct but appropriately packaged for its intended audience and purpose.
“Write in a confident, executive-briefing style — data-driven but accessible. Use bullet points for action items, avoid jargon, and lead each section with the business impact before the technical detail.”
E — Experiment
Request multiple variations, alternative approaches, or different angles on the same problem. The Experiment dimension is what makes CRISPE uniquely powerful — instead of accepting a single output, you ask the AI to explore the solution space. This leverages the probabilistic nature of language models and gives you genuine creative options to evaluate, combine, or iterate upon.
“Provide three distinct strategy variants: one focused on pricing adjustments, one on product feature improvements, and one on customer success program enhancements. For each variant, include the expected churn reduction and estimated implementation cost.”
See the Difference
Why a structured CRISPE prompt outperforms a basic request
Basic Prompt
Help me reduce customer churn for my SaaS product.
Here are some ways to reduce churn: improve onboarding, offer discounts, send re-engagement emails, improve customer support, and add new features customers want.
CRISPE Prompt
Capacity/Role: Senior data scientist, churn analytics expert
Insight: 50K users, 4.2% monthly churn (up from 3.1%), competitors launched free tiers, support tickets up 30%
Statement: 90-day churn reduction strategy with metrics and roadmap
Personality: Executive-briefing style, data-driven, accessible
Experiment: Three variants: pricing, product features, customer success
Three distinct strategies, each with projected churn reduction percentages, implementation timelines, resource requirements, and ROI estimates — all written in executive-ready language with clear prioritization.
Natural Language Works Too
While structured frameworks and contextual labels are powerful tools, LLMs are exceptionally good at understanding natural language. As long as your prompt contains the actual contextual information needed to create, answer, or deliver the response you’re looking for — the who, what, why, and constraints — the AI can produce complete and accurate results whether you use a formal framework or plain conversational language. But even in 2026, with the best prompts, verifying AI output is always a necessary step.
CRISPE in Action
See how each dimension shapes the AI’s response
Capacity/Role: You are a senior content marketing strategist with 12 years of experience in B2B SaaS, specializing in demand generation and thought leadership campaigns.
Insight: Our company sells project management software to mid-market companies (200–2,000 employees). We’re entering the European market for the first time. Our current content performs well in the US but hasn’t been adapted for European audiences. Budget is $15K/month for content production.
Statement: Create a 3-month content calendar for our European market entry, covering blog posts, LinkedIn content, webinars, and email sequences.
Personality: Write in a strategic, consultative tone. Be specific with recommendations but explain the reasoning behind each choice. Use a format suitable for presenting to the CMO.
Experiment: Provide two calendar variants: one focused on localized thought leadership and one focused on competitive displacement campaigns.
The Capacity grounds the AI in the right expertise domain. The Insight provides specific business constraints (budget, market size, current situation) that shape realistic recommendations. The Statement defines a concrete deliverable. The Personality ensures the output is boardroom-ready. The Experiment gives the marketing team two distinct strategic directions to evaluate — rather than committing to one approach blindly.
Capacity/Role: You are a cloud infrastructure architect with deep expertise in AWS, microservices patterns, and high-availability system design for financial services applications.
Insight: We run a payment processing platform handling 2 million transactions per day. Current architecture is a monolith on AWS EC2 instances. We’ve been experiencing scaling issues during peak hours (3x normal load). Our team has 8 backend engineers, most with limited container orchestration experience. Compliance requires PCI DSS Level 1.
Statement: Design a migration strategy from our monolith to microservices that addresses our scaling issues while maintaining PCI compliance and accounting for our team’s skill gaps.
Personality: Be direct and technically precise. Use architecture decision records (ADR) format where appropriate. Flag risks explicitly and rate them by severity.
Experiment: Present two migration paths: a conservative strangler-fig approach and an aggressive parallel-build approach. Compare timelines, risk profiles, and team ramp-up requirements for each.
The Insight dimension is critical here — transaction volume, team size, skill gaps, and compliance requirements all constrain the solution space dramatically. Without these details, the AI would produce generic microservices advice. The Experiment component ensures the team gets both a safe path and an ambitious path, with honest trade-off analysis for each.
Capacity/Role: You are an instructional designer specializing in adult learning theory and corporate training programs, with particular expertise in designing courses for technical professionals transitioning to leadership roles.
Insight: Our engineering department is promoting 12 senior engineers to team lead positions over the next quarter. These individuals are technically excellent but have no formal management training. The company values servant leadership and psychological safety. Training budget allows for 4 half-day workshops plus ongoing monthly sessions.
Statement: Design a leadership development curriculum covering the transition from individual contributor to team lead, including workshop outlines, practical exercises, and assessment criteria.
Personality: Write in an encouraging, practical tone. Avoid corporate buzzwords. Use concrete scenarios engineers would recognize. Structure content as a facilitator’s guide with timing notes.
Experiment: Offer two curriculum structures: one built around progressive skill-building (each workshop builds on the previous) and one built around common failure scenarios (each workshop addresses a specific leadership challenge new managers face).
The Personality dimension transforms this from a generic training outline into something usable. By requesting engineer-friendly language and facilitator-ready formatting, the output arrives almost ready to deliver. The Experiment dimension is especially valuable for curriculum design — the progressive structure works for systematic learners, while the scenario-based structure works for those who learn best from real-world problems.
When to Use CRISPE
Best for complex tasks requiring expertise, context, and exploration
Perfect For
Business strategies, market analyses, and organizational plans where multiple viable approaches exist and need side-by-side comparison.
Tasks requiring deep domain expertise — legal analysis, medical reasoning, financial modeling — where the AI’s assumed capacity directly impacts output quality.
Design briefs, campaign concepts, or product ideas where generating multiple distinct variations is more valuable than a single “best” answer.
Reports, presentations, and documents where tone, format, and audience-appropriateness matter as much as the content itself.
Skip It When
Simple questions with single correct answers — “What year was Python released?” — don’t benefit from six dimensions of context.
When you’re in a fast back-and-forth conversation refining an idea, the overhead of a full CRISPE prompt slows the creative flow.
When you know exactly what format you want and don’t need variations — a simpler framework like CRISP or COSTAR will do the job with less effort.
Use Cases
Where CRISPE delivers the most value
Market Research
Define the analyst role, supply competitive intelligence, request market assessment, set the reporting tone, and explore bullish vs. bearish scenarios side by side.
Proposal Writing
Set the AI as a proposal specialist, provide project context and client preferences, request a tailored proposal, and generate versions for different budget tiers.
Product Strategy
Assign product leadership expertise, share usage data and competitive landscape, request a feature roadmap, and experiment with growth-led vs. retention-led strategies.
Training Programs
Cast the AI as an instructional designer, provide learner demographics and constraints, request curriculum design, and explore workshop vs. self-paced delivery models.
Risk Assessment
Establish compliance expertise, supply regulatory context and operational data, request a risk matrix, and generate conservative vs. aggressive risk tolerance scenarios.
Data Analysis
Define the analyst’s statistical specialization, provide dataset context and business questions, request insight reports, and explore different analytical lenses on the same data.
Where CRISPE Fits
CRISPE bridges simple role prompting and full multi-agent workflows
CRISPE’s Experiment dimension sets it apart from other structured frameworks. While CRISP and COSTAR produce a single optimized response, CRISPE intentionally generates multiple approaches. This is especially valuable in early-stage planning, creative work, and strategic decision-making — situations where the “right” answer isn’t obvious and exploring the option space is itself the goal.
Related Techniques & Frameworks
Explore complementary structured approaches
Build Your CRISPE Prompt
Use our interactive tools to construct structured CRISPE prompts or find the right framework for your specific task.