The RISEN Framework: Write Prompts That Work Every Single Time

Why Your Prompts Are Inconsistent

You've had the experience: you write a prompt, get a perfect response, then copy a slightly different version next week and get something completely off. The output is too long. The tone is wrong. It misses the key point. You rewrite it three times and still don't get what you need.

This isn't AI being unpredictable. This is vague instructions producing vague results. AI is a brilliant system — but it needs structure to perform at its best. Without structure, it fills the gaps with its own assumptions, and those assumptions are rarely aligned with what you actually need.

The RISEN framework solves this. It gives every prompt a clear, complete structure that removes ambiguity and delivers consistent, professional outputs every time.

💡 What Is RISEN?

RISEN stands for Role, Instructions, Steps, End Goal, Narrowing. It is a 5-part prompt structure developed for business professionals who need consistent, repeatable AI outputs. Each element eliminates a specific type of ambiguity that causes inconsistent results.

Before RISEN vs After RISEN

Here's the same request — written without structure versus written with RISEN. The difference in output quality is dramatic:

❌ Without RISEN (Vague)

"Write an email about the invoice that's late."

Result: Generic email. Wrong tone. Missing key details. Needs complete rewrite.

✅ With RISEN (Structured)

"Role: Senior finance professional. Task: Write payment follow-up to [VENDOR] re invoice [NO] for [AMOUNT]. Steps: 1) Acknowledge overdue status 2) State payment timeline 3) Request confirmation. End goal: Vendor confirms payment date. Narrowing: Professional tone. Max 120 words. Include subject line."

Result: Professional, on-tone, correctly structured email. Ready to send in 30 seconds.

The 5 Elements of RISEN — Explained

Role

Tell AI who it is before asking it to do anything

This is the single most impactful element. When you assign a role, AI calibrates its entire response — vocabulary, tone, depth, and perspective — to match that identity. A "senior finance analyst" writes very differently to a "customer service agent" even given identical instructions.

Example

"You are a senior Accounts Executive with 8 years of experience in P2P and O2C operations at a professional services firm."

Instructions

Tell AI exactly what to do — not what you want to talk about

Most people describe the situation when they should be giving a direct task instruction. "I have an invoice dispute" is a situation. "Write a professional response to a customer invoice dispute" is an instruction. Be direct, specific, and action-oriented.

Example

"Write a professional email response to customer [NAME] who has disputed invoice [NUMBER] for [AMOUNT] citing [REASON]."

Steps

Give AI a process to follow — not just a destination

Without steps, AI decides its own structure, which often doesn't match what you need. When you define the steps, you control the flow of the output. This is especially important for reports, analyses, and complex communications where sequence matters.

Example

"Step 1: Acknowledge the dispute professionally. Step 2: State our position clearly. Step 3: Propose a resolution path. Step 4: End with a clear call to action."

End Goal

Tell AI what success looks like for this output

The end goal tells AI what the output will be used for and what reaction you want it to produce. This changes how AI frames its language — a response designed to "get customer to pay immediately" will be written differently to one designed to "preserve the client relationship while recovering payment."

Example

"The goal is for the customer to acknowledge the dispute is resolved and confirm payment within 5 working days — while preserving the business relationship."

Narrowing

Add constraints that eliminate all the outputs you don't want

Narrowing is the most commonly skipped element — and the most impactful one for consistency. Without constraints, AI will choose its own length, format, vocabulary, and tone. Narrowing removes that freedom and forces AI to produce exactly what you need. Always include at least word count, format, and tone.

Example

"Max 150 words. Include subject line. Professional but empathetic tone. No legal language. End with a specific deadline request."

A Full RISEN Prompt in Action

Here's a complete RISEN prompt for one of the most common finance tasks — a month-end closing report to management:

Full RISEN Prompt Example Ready to Copy

Role:
You are a senior Accounts Executive reporting to the Finance Manager
at a mid-sized professional services company in India.

Instructions:
Write a month-end closing status update for [MONTH] [YEAR]
to share with the Finance Manager and CFO.

Steps:
1. Summarise what has been completed this month-end
2. List what is still pending with expected completion dates
3. Flag any blockers or dependencies
4. Provide key financial metrics snapshot
5. State the overall closing readiness status

End Goal:
Leadership should be able to read this in 90 seconds and know
exactly where month-end stands, what risks exist, and what
action if any is needed from them.

Narrowing:
Max 250 words. Use H3 headers for each section.
Professional and direct tone — no filler sentences.
RAG status required (Red / Amber / Green) for overall closing.
Ready to paste into an email to CFO without editing.

--- Fill in these variables before using ---
Completed items: [LIST]
Pending items + ETA: [LIST]
Blockers: [LIST or "None"]
Key metrics: AR=[AMT] | AP=[AMT] | Unreconciled=[AMT]

✅ Pro Tip — You Don't Always Need All 5

For simple tasks, R + I + N is enough. Start with Role and Instructions always. Add Steps when sequence matters. Add End Goal when the purpose changes the tone. Add Narrowing (always) to control format and length. Test and simplify based on results.

Get 10 Pre-Built RISEN Templates Free

The Data Foundry AI Starter Playbook includes 10 fully structured RISEN templates across finance, operations, and management tasks.

Download Free Playbook Shop on Etsy

When to Use Each Element

Here's a quick reference guide for when each RISEN element has the most impact:

Element	Always Use?	Especially Important For
R — Role	Yes	Any professional communication, reports, analysis
I — Instructions	Yes	Every single prompt — this is your task definition
S — Steps	Sometimes	Multi-part documents, reports, structured analysis
E — End Goal	Sometimes	Persuasive emails, negotiations, client communications
N — Narrowing	Yes	Everything — always define length, format, and tone

Common Mistakes When Using RISEN

Being vague in the Role. "You are a professional" gives AI nothing useful. "You are a senior finance manager with expertise in P2P operations writing for a CFO audience" gives AI everything it needs.
Skipping Narrowing. This is the most commonly skipped element and the one that causes the most inconsistency. Always define at minimum: word count, format, and tone.
Too many steps. Keep Steps to 3–5 clear items maximum. More than that and AI starts producing a mechanical checklist rather than a coherent document.
Not saving prompts that work. When a RISEN prompt gives you an excellent output — save it immediately. Build your prompt library one successful RISEN prompt at a time.

"The goal isn't to write better prompts on the fly — it's to build a library of RISEN prompts so good that you never need to improvise again."

— Selvi, Founder · Data Foundry

Your RISEN Practice Exercise

The best way to learn RISEN is to use it right now on a real task. Here's a quick exercise:

Pick one task you do repeatedly — a weekly report, a type of email, a summary you write.
Write a RISEN prompt for that task using all 5 elements. Take 5 minutes.
Test it 3 times in Claude or ChatGPT with different variable inputs.
Score each output 1–5. If you get 4+ consistently — add it to your library as a permanent template.

Done this once? You now have a template that will save you time every week for years. Repeat this for your top 10 tasks and you have a complete professional prompt library.