RISEN Prompt Framework Examples for Support & Analysis

RISEN prompt framework examples give support and analysis teams labeled blocks they can paste into ChatGPT, Claude, or Gemini without inventing structure each ticket. RISEN means Role, Instructions, Steps, End-goal, and Narrowing. This page is a paste library for refunds, escalations, macro drafts, ticket clustering, root-cause notes, and research briefs you can run the same day. Soft path when rough notes need labels: https://promptmake.net/frameworks/risen. You leave with eight full examples, solo and team edit passes, common failure fixes, and an FAQ. Field-by-field theory and the multi-framework hub ship elsewhere. Stay here for support and analysis copy you can paste today.

Who these RISEN prompt framework examples serve

Support leads who want consistent tone across tier-1 and tier-2 replies. Analysts who turn messy ticket dumps into ranked themes. Ops writers who draft macros that still need human review before send. Managers who need a weekly severity brief without invented metrics.

RISEN fits these jobs because sequence and done-state matter. A refund reply needs acknowledge, policy, next step. A root-cause note needs evidence, hypothesis, open questions. RTF is too light when Narrowing must block legal promises. CO-STAR shines on tone-heavy marketing. Soft draft path: https://promptmake.net/frameworks/risen.

Models: GPT-5.6 Sol, Claude Fable 5, Claude Opus 5, and Gemini 3.5 Flash all follow the five headers when Context sits after the block. Skip "think step by step" filler on reasoning-class models. Put clear End-goal and Narrowing instead.

PromptMake generates labeled text only. It does not send email, close tickets, or store a vault unless your org adds that separately. Guest and free quotas apply per text tool path. Confirm limits on promptmake.net.

Quick RISEN reminder before you paste

Role sets voice and reader. Instructions state the main verb. Steps order the work when order changes quality. End-goal names the finished artifact. Narrowing lists length, honesty, tone bans, and exclusions. Context goes last with the ticket, policy, or dataset.

Copy this blank shell when you adapt any example below:

Role: [voice + audience]

Instructions: [verb + object]

Steps:

1. [milestone]
2. [milestone]
3. [milestone]

End-goal: [artifact + format]

Narrowing: [length, exclusions, honesty]

Context: [paste source]

Keep Role under three lines. Keep Steps between three and seven. If Narrowing is empty, the model will sprawl. Soft bootstrap: https://promptmake.net/frameworks/risen.

Support RISEN prompt framework examples

Support examples below assume a human still reviews before send. Narrowing blocks legal admissions, invented SLAs, and medical or financial advice outside policy. Replace product nouns with your own. Keep policy text in Context so the model cannot invent refund windows.

Run each example once on your default chat model. Score against End-goal only. If length drifts, tighten Narrowing before you rewrite Role. Save winners in your helpdesk macro library with the model name and date.

These blocks target risen prompt framework practice for customer-facing work. They are not CO-STAR tone drills and not the general RISEN deep dive. Read one example aloud before paste so empty Steps or vague End-goals jump out.

Example 1: duplicate charge refund reply

Role: Tier-1 support agent for a SaaS billing product. Calm, plain language. Reader is a frustrated customer on mobile.

Instructions: Draft a reply about a possible duplicate charge using only the ticket and policy in Context.

Steps:

1. Acknowledge the reported duplicate charge and thank the customer for the details.
2. Restate refund eligibility rules from Context without adding new timelines.
3. Give the next verification step and what the customer should reply with.

End-goal: Email with Subject line plus body under 140 words. Sections flow as short paragraphs. Sign-off: Support Team.

Narrowing: No admission of company fault unless Context confirms it. No invented refund dates. No legal threats. US English. No emojis.

Context: [paste ticket thread + refund policy excerpt]

Example 2: outage status update to affected users

Role: Status-page writer for an infrastructure product. Direct, non-alarmist.

Instructions: Write a status update for a partial API outage based only on the incident notes in Context.

Steps:

1. State what is broken in one sentence.
2. State current mitigation from Context.
3. State next update window using only times present in Context.

End-goal: Status blurb under 90 words plus a 140-character social variant.

Narrowing: No root cause guesses. No ETA unless Context includes one. Mark missing times as UPDATE PENDING. No blame on vendors by name unless Context names them.

Context: [paste incident commander notes]

Example 3: escalation summary for tier-2

Role: Tier-1 agent handing a case to tier-2. Precise, no fluff.

Instructions: Summarize the ticket for a specialist who has sixty seconds to read.

Steps:

1. List customer goal in one line.
2. List troubleshooting already tried from Context.
3. List open questions and attachments needed.

End-goal: Markdown with H2 sections Goal | Tried | Open questions | Risk. Max 180 words.

Narrowing: No new troubleshooting ideas. No tone judgments about the customer. Quote error codes exactly. If a step is missing from Context, write NOT IN THREAD.

Context: [paste full ticket]

Example 4: help center macro from a resolved thread

Role: Knowledge editor for a support org. Neutral how-to voice.

Instructions: Turn a resolved ticket into a reusable help article outline.

Steps:

1. Extract the problem statement customers would search.
2. List verified fix steps from Context only.
3. Add one "still stuck" escalation line.

End-goal: Markdown article draft with Title, Summary (40 words), Steps numbered, Escalation. Max 250 words.

Narrowing: No undocumented workarounds. No screenshots claimed unless Context describes them. No product promises beyond Context. US English.

Context: [paste resolved thread + product version]

Analysis RISEN prompt framework examples

Analysis examples turn messy text into ranked findings. They belong in weekly ops reviews, QA sampling, and research assists for product managers. End-goal should name tables or bullet ranks so the model stops at a usable artifact. Support macros above focus on outbound customer language. The blocks below focus on internal decision artifacts your team can archive.

Honesty rules matter more here than clever Role titles. Force UNVERIFIED tags when evidence is thin. Ban invented percentages. Soft path when you start from a rough brief: https://promptmake.net/frameworks/risen.

Paste large transcripts after the RISEN block. Instructions first, data second. Reasoning models still need a hard stop definition or they write essays when you wanted a five-row table. Keep one primary analysis job per prompt so Steps stay short enough to finish.

Example 5: ticket theme clustering

Role: Support operations analyst writing for a support manager.

Instructions: Cluster the pasted tickets into themes and rank them by volume proxies available in Context.

Steps:

1. Skim all tickets and propose five to eight theme labels.
2. Assign each ticket ID to one primary theme.
3. Rank themes by count and note one sample quote per theme.

End-goal: Markdown table with columns Theme | Count | Sample quote | Suggested owner team. Max eight rows.

Narrowing: No themes without at least one ticket ID. No invented CSAT scores. If volume is unclear, rank by listed order and tag COUNT UNKNOWN. Neutral wording.

Context: [paste ticket export]

Example 6: root-cause hypotheses from logs

Role: On-call analyst drafting a pre-mortem note for engineering.

Instructions: Propose ranked root-cause hypotheses from the logs and timeline in Context.

Steps:

1. Restate the user-visible symptom in one sentence.
2. List timeline facts with timestamps from Context only.
3. Rank three hypotheses with supporting and contradicting evidence.

End-goal: Markdown with Symptom | Timeline bullets | Hypotheses table (Hypothesis, Support, Against, Confidence high/med/low).

Narrowing: No single root cause declared. No blame language. Tag speculative lines UNVERIFIED. Max 220 words outside the table.

Context: [paste logs + timeline]

Example 7: competitor support policy comparison

Role: Product operations analyst writing for a VP of Support.

Instructions: Compare refund and SLA language across competitors using only the pasted policy excerpts.

Steps:

1. List comparison criteria present in Context.
2. Fill a table per competitor with cells drawn from excerpts.
3. Write a 60-word implication note for our policy review.

End-goal: One-page brief. Criteria list, comparison table, implications paragraph.

Narrowing: Mark missing cells NEEDS RESEARCH. No invented pricing or SLA hours. No trash talk. Cite competitor names only as given.

Context: [paste policy excerpts]

Example 8: weekly severity brief from QA samples

Role: QA lead writing a Monday brief for support leadership.

Instructions: Summarize severity patterns from the QA sample sheet in Context.

Steps:

1. Count tickets by severity labels present in Context.
2. List top three failure modes with example ticket IDs.
3. Propose two process experiments for next week tied to those modes.

End-goal: Email under 200 words with sections Counts | Failure modes | Experiments.

Narrowing: No severity relabeling. No experiments that require new headcount. No customer names. If a count is missing, write DATA GAP.

Context: [paste QA sheet]

How to adapt these RISEN prompt framework examples

Swap product nouns first. Then rewrite Narrowing to match your legal macros. Then adjust End-goal length to your channel: chat widgets need shorter bodies than email. Keep Steps stable so reviewers know where to look.

For regulated industries, move compliance lines into Narrowing and keep a human approval gate. The model drafts. Your agent sends. Soft path to bootstrap a new job from notes: https://promptmake.net/frameworks/risen.

Version templates monthly. Delete examples that embed old refund windows. Store model name with each win because GPT-5.6 Sol and Claude Fable 5 can drift on length even with the same Narrowing.

Solo edit pass

Read End-goal first after a run. If the artifact shape is wrong, fix End-goal and Narrowing before Role. Cut Steps that repeat Instructions. Move Context below the five fields if you pasted data first by habit.

Team edit pass

Assign one owner per macro. Review in a thirty-minute monthly meeting. Reject templates that invent policy. Link new hires to the RISEN deep dive for field definitions and keep this page as the support and analysis paste shelf.

Common mistakes on support and analysis prompts

Mistake: Role is a long persona novel. Fix: audience and tone in three lines.

Mistake: Instructions ask for reply plus analysis plus tweet. Fix: split prompts.

Mistake: End-goal says "helpful summary." Fix: name sections and word caps.

Mistake: Empty Narrowing on refund work. Fix: ban admissions and invented dates.

Mistake: Context above RISEN so Steps get ignored. Fix: contract first, paste second.

Mistake: Mixing CO-STAR and RISEN headers. Fix: one skeleton per job.

Repair path: empty the chat, paste one example, swap Context, score against End-goal once. Soft draft path: https://promptmake.net/frameworks/risen.

FAQ

What are RISEN prompt framework examples?

They are paste-ready prompts that use Role, Instructions, Steps, End-goal, and Narrowing labels. This page focuses on support replies and analysis briefs. Soft generator path: https://promptmake.net/frameworks/risen. Edit Narrowing for your policy before production send.

When should support teams use the risen prompt framework?

Use RISEN when sequence and done-state drive quality: refunds, escalations, status updates, ticket clustering, and root-cause notes. Use RTF for tiny one-line rewrites. Use CO-STAR when brand tone is the main risk on marketing copy.

How is this different from risen-framework-prompting?

Risen-framework-prompting is the field-by-field deep dive with mixed job types. This article is a support and analysis paste library. Read the deep dive to learn fields. Use this page when you need ticket-ready blocks.

Can I use these examples with Claude and ChatGPT?

Yes. Paste the five fields, then Context. Claude Fable 5, GPT-5.6 Sol, and Gemini 3.5 Flash follow the structure. Reasoning-class models still need clear End-goal and Narrowing. Skip chain-of-thought filler.

Where does ticket text go in a RISEN prompt?

Put ticket text, policies, and logs in Context after the five RISEN fields. Instructions belong first so the model sees the contract before the data. Long pastes above the headers often cause skipped Steps.

How do I start on PromptMake for free?

Open https://promptmake.net/frameworks/risen, describe a support or analysis job in plain language, and edit the labeled output. Guest and free account limits apply per tool path. Confirm quotas on promptmake.net before a team workshop.

What belongs in Narrowing for support replies?

Length caps, banned admissions, no invented SLAs, source-only policy, language and emoji rules, and escalation limits. If a rule prevents a risky send, it belongs in Narrowing. If it defines the deliverable shape, it belongs in End-goal.