FrameworkGuidebeginner12 min read

The RISEN Prompt Framework: A Complete Guide

Structure complex prompts with Role, Instructions, Steps, End Goal, and Narrowing

Why RISEN Works

AI is powerful, but it does not read minds. When you write a vague prompt, the AI fills in the gaps with assumptions. Sometimes those assumptions are right. More often, they are not.

The RISEN framework solves this by giving you a repeatable structure that tells AI exactly who to be, what to do, how to do it, what success looks like, and what to avoid. It is especially effective for complex, multi-step tasks where you need the AI to follow a process, not just generate text.

Where other frameworks focus on communication style and audience, RISEN is action-oriented. It breaks work into steps with clear end goals. Think of it as giving AI a project brief with a built-in checklist.

Insight

RISEN excels at tasks that require a process: onboarding plans, project proposals, technical analyses, training materials. Any task where the “how” matters as much as the “what.”

What is RISEN?

RISEN is an acronym for five prompt elements that together create structured, process-driven AI instructions. Each letter represents a component that shapes how the AI approaches your task:

1

Role

The persona or expert identity the AI should adopt. This shapes its knowledge base, perspective, and approach.
2

Instructions

The specific task or action you want the AI to perform. Clear, direct instructions eliminate ambiguity.
3

Steps

The sequence of actions or phases the AI should follow. Breaking work into steps produces more thorough, organized output.
4

End Goal

The desired outcome or success criteria. Defining the destination helps the AI navigate toward the right result.
5

Narrowing

Constraints, boundaries, and exclusions that focus the output. Narrowing prevents scope creep and keeps responses relevant.

The power of RISEN is in the combination. Role sets the expertise. Instructions define the task. Steps provide the process. End Goal defines success. Narrowing keeps it focused. Together, they eliminate the guesswork that leads to generic AI outputs.

R - Role

Role tells the AI who it should be. When you assign a role, the AI draws on knowledge, vocabulary, and judgment associated with that persona. A “senior data analyst” approaches a dataset differently than a “business journalist.”

The more specific the role, the better. “Marketing consultant” is good. “B2B SaaS marketing consultant with 10 years of experience in developer tools” is much better. Specificity activates more relevant knowledge within the model.

Role Example
ROLE:
You are a senior operations manager with 15 years of experience in logistics and supply chain optimization for mid-size e-commerce companies. You specialize in warehouse workflow design and have implemented process improvements that reduced fulfillment times by 30-40% at three previous companies.
  • Name the expertise: “Senior financial analyst specializing in SaaS metrics” not just “analyst”
  • Add experience level: Years of experience or seniority level anchors the depth of response
  • Include domain: The industry or field matters. A healthcare PM thinks differently than a tech PM
  • Mention relevant achievements: Past accomplishments signal the caliber of expertise you expect
  • Match role to task: The role should be the type of expert who would naturally handle your request

Pro Tip

Test different roles for the same task. A “startup founder” and a “management consultant” will give you different perspectives on the same business problem. Sometimes the unexpected role produces the best insight.

I - Instructions

Instructions are the core task definition. This is where you tell the AI exactly what you need it to do. Strong instructions use action verbs and leave no room for misinterpretation.

The difference between weak and strong instructions is specificity. “Help me with onboarding” gives AI too much freedom. “Create a 30-day onboarding plan for new software engineers joining a 50-person startup” gives AI exactly what to build.

Before
Help me improve our warehouse operations.
After
INSTRUCTIONS:
Create a detailed warehouse process audit for our current order fulfillment workflow. Identify the top 5 bottlenecks causing delays, rank them by impact on fulfillment time, and propose a specific solution for each one. Include estimated time savings and implementation difficulty for each solution.
  • Start with an action verb: Create, analyze, design, write, evaluate, compare, develop
  • Specify the deliverable: What exact artifact should the AI produce?
  • Set scope boundaries: How many items? What time period? What depth?
  • Include quality criteria: “Rank by impact” or “include supporting data”
  • Be explicit about format needs: If you need a table, numbered list, or specific structure, say so here

S - Steps

Steps are what make RISEN unique among prompt frameworks. Instead of leaving the AI to figure out how to approach a complex task, you define the sequence explicitly. This produces more thorough, logically organized output.

Think of steps as giving the AI a recipe. Without steps, it might skip important phases or tackle things in the wrong order. With steps, you control the flow and ensure nothing gets missed.

Steps Example
STEPS:
1. First, analyze the current fulfillment process from order receipt to shipment, mapping each stage and its average completion time.
2. Next, identify where delays occur by comparing expected vs. actual times at each stage.
3. Then, categorize the bottlenecks: staffing, technology, layout, process design, or external factors.
4. For each bottleneck, research and propose one primary solution and one backup alternative.
5. Finally, create an implementation timeline that sequences the fixes from quickest wins to longer-term projects.
  • Number your steps: Sequential numbering makes the order unambiguous
  • Use transition words: “First,” “Next,” “Then,” “Finally” reinforce the sequence
  • Keep steps independent: Each step should produce a clear intermediate output
  • Order by logic, not importance: Put research before recommendations, analysis before conclusions
  • Limit to 3-7 steps: Too many steps create confusion; too few lose the benefit of structure

Insight

Steps are what set RISEN apart. They force the AI to think methodically rather than jumping straight to conclusions. The result is more nuanced, comprehensive output that mirrors how an expert would actually approach the problem.

E - End Goal

The End Goal defines what success looks like. It tells the AI the purpose behind the task, not just the task itself. When AI understands the destination, it makes better decisions at every step along the way.

An End Goal answers: What will you do with this output? What decision will it inform? What outcome are you working toward? This context helps the AI calibrate depth, emphasis, and practicality.

End Goal Example
END GOAL:
The final deliverable should be a presentation-ready report that I can share with the VP of Operations to secure budget approval for warehouse improvements. The report needs to clearly show ROI for each proposed change, with projected cost savings over 6 and 12 months. It should be compelling enough to justify a $50,000-$75,000 investment in process improvements.
  • State the purpose: Why do you need this? What decision does it support?
  • Name the audience: Who will see the final output? This shapes polish and tone
  • Define “done”: What does the output need to contain to be complete?
  • Include success metrics: How will you measure whether the output achieved its goal?
  • Be honest about stakes: A quick internal note requires different effort than a board presentation

Pro Tip

The End Goal is your most underused lever. Most people skip it, but stating “this needs to convince a skeptical CFO” produces noticeably different output than “this is for my own reference.”

N - Narrowing

Narrowing is about constraints and boundaries. It tells the AI what to exclude, what limits to respect, and what scope to stay within. Without narrowing, AI tends to be broad and generic. With narrowing, it becomes focused and relevant.

Think of narrowing as guardrails. You are not limiting the AI’s capability - you are directing its energy toward what matters most for your specific situation.

Narrowing Example
NARROWING:
- Focus only on the pick-and-pack stage of fulfillment; exclude receiving, storage, and shipping carrier selection.
- Solutions must be achievable within 90 days and with existing staff (no new hires).
- Do not recommend any solution requiring more than $20,000 upfront investment per improvement.
- Exclude automation solutions that require custom software development.
- Keep the report under 2,000 words; use tables and bullet points rather than long paragraphs.
  • Set scope limits: Which parts of the topic to include and exclude
  • Define constraints: Budget, timeline, team size, tools available
  • Specify exclusions: “Do not include...” is powerful for preventing irrelevant content
  • Set length limits: Word counts, number of items, page limits
  • Add format constraints: “Use tables instead of paragraphs” or “no more than 3 levels of nesting”

Warning

Narrowing is not about being restrictive for its own sake. Every constraint should serve the End Goal. If a constraint does not make the output more useful, remove it.

RISEN in Action

See the difference between a prompt without structure and one that uses the full RISEN framework. The task is the same. The results are not.

Before
Help me create an onboarding plan for new engineers.
After
ROLE:
You are a VP of Engineering with 12 years of experience building and scaling engineering teams at high-growth startups (Series A through C). You have onboarded over 80 engineers and refined your onboarding process through three different companies.

INSTRUCTIONS:
Create a 30-day onboarding plan for new software engineers joining a 50-person B2B SaaS startup. The plan should cover technical setup, codebase orientation, team integration, and first meaningful contribution. Structure it as a week-by-week schedule with specific daily activities.

STEPS:
1. Start with pre-arrival preparation: what should be ready before day one.
2. Design Week 1 around environment setup, tool access, and meeting key team members.
3. Design Week 2 around codebase exploration, architecture walkthroughs, and first small PR.
4. Design Week 3 around working on a starter project with a mentor, attending sprint ceremonies.
5. Design Week 4 around independent work on a real feature, with a 30-day review checkpoint.

END GOAL:
This plan will become our standard engineering onboarding template. It needs to reduce time-to-first-commit from the current average of 14 days to under 7 days, and help new engineers feel productive and integrated within their first month. The hiring manager and HR lead will review and approve it.

NARROWING:
- Focus on backend engineers working with Node.js and PostgreSQL; exclude frontend-specific onboarding.
- Assume remote-first team across US time zones.
- Do not include HR paperwork or benefits enrollment (HR handles that separately).
- Keep each day's activities to 4-6 hours of structured time; leave room for self-directed learning.
- Format as a table with columns: Day, Focus Area, Activities, Owner, Success Criteria.

Success

The RISEN prompt takes a few minutes to write. But it replaces multiple rounds of “that’s not quite right, can you adjust...” and produces output you can actually use.

RISEN vs COSTAR: When to Use Which

RISEN and COSTAR are both excellent prompt frameworks, but they solve different problems. They are complementary tools, not competitors. Choosing the right one depends on your task.

Use RISEN when your task involves a process, requires multiple steps, or needs the AI to follow a specific methodology. RISEN excels at complex, action-oriented work where the “how” matters as much as the “what.”

Use COSTAR when your task is primarily about communication: writing content, crafting messages, or producing output where style, tone, and audience shape the quality. COSTAR excels at content that needs to land with a specific reader.

AspectRISENCOSTAR
FocusProcess and methodologyCommunication and style
Best forMulti-step tasks, analyses, plansContent writing, emails, messaging
Unique strengthSteps and End Goal drive structured outputStyle, Tone, and Audience shape voice
Example tasksOnboarding plans, audits, proposals, researchBlog posts, sales emails, social copy
Handles constraintsNarrowing (explicit exclusions and limits)Response Format (output structure)

Insight

You can combine elements from both frameworks. Use RISEN’s Steps with COSTAR’s Audience and Tone for complex writing tasks that need both process and voice. Frameworks are tools, not rules.

Quick Reference Cheatsheet

Use this cheatsheet when building your next prompt. Not every prompt needs all five elements, but check each one to decide what to include.

ElementQuestion to AskExample
R - RoleWhat expert should tackle this?Senior operations manager, 15 years in logistics
I - InstructionsWhat exactly should AI produce?Create a warehouse process audit with top 5 bottlenecks
S - StepsWhat sequence should AI follow?Map process, identify delays, categorize, propose fixes
E - End GoalWhat does success look like?Presentation-ready report showing ROI for VP approval
N - NarrowingWhat boundaries and exclusions apply?Pick-and-pack only, under $20K per fix, 90-day timeline
RISEN Template
RISEN Template - Copy and Fill In:

ROLE:
[The expert persona AI should adopt: title, experience level, domain expertise]

INSTRUCTIONS:
[The specific task with action verb, deliverable, and scope]

STEPS:
1. [First action or phase]
2. [Second action or phase]
3. [Third action or phase]
4. [Fourth action or phase]
5. [Final action or phase]

END GOAL:
[What success looks like: purpose, audience for the output, how it will be used]

NARROWING:
- [Scope limitation or exclusion]
- [Budget, timeline, or resource constraint]
- [Format or length requirement]
- [Any other boundary that focuses the output]

Next Steps

You now have the RISEN framework. Try it on your next complex task and compare the results to your usual prompting approach. The difference is usually obvious on the first try.

Want to explore more frameworks? The COSTAR method guide covers a complementary framework that excels at content and communication tasks. Together, RISEN and COSTAR handle the vast majority of prompting needs.

Or skip the manual work entirely. AskSmarter.ai’s prompt builder asks you smart questions and automatically applies frameworks like RISEN and COSTAR behind the scenes. You get structured, optimized prompts without memorizing any acronyms.

Build RISEN prompts automatically

Answer a few questions about your task. AskSmarter applies RISEN, COSTAR, and other proven frameworks to create prompts that get results. No prompt engineering expertise required.

Start building free