Prompt Chain: Research Brief to Point-of-View Memo
A three-step prompt chain for turning raw research into a sharper point-of-view memo without skipping the reasoning in the middle.
Use cases
Research & Analysis, Strategy & Planning
Platforms
Claude, GPT, Model-Agnostic
Jump to a section
The resource
Copy and adapt. Do not paste blind.
Stage 1: Extract the core facts, tensions, patterns, and open questions from the research brief. Keep facts, interpretations, and assumptions separate.
Stage 2: Generate 2-3 possible points of view the memo could argue. For each one, include:
- core argument
- strongest supporting evidence
- likely pushback
- where evidence is still thin
Stage 3: Write the final memo using the selected point of view.
Memo structure:
1. Thesis
2. What the research shows
3. What it likely means
4. Implications
5. Risks and caveats
6. Recommended next movesWhen to Use This
Use this when you have research inputs but still need to decide what argument or interpretation should come out the other end.
It suits strategists, operators, founders, and consultants who need something more useful than a neutral summary but do not want the model to jump straight into hot takes.
Why It Works
The chain forces an intermediate step between raw synthesis and final memo writing. That is where the useful judgement happens.
By making the model propose multiple possible points of view before drafting, you reduce the risk of getting one generic argument and mistaking it for the best one.
How to Customise
If your memo has to speak to a specific audience, add that constraint before stage three. A board memo, founder memo, and client memo do not sound the same.
You can also require a tighter evidence table in stage one if the input material is messy or politically sensitive.
Limitations
This improves reasoning flow, but it does not validate source quality or resolve contradictory evidence on its own.
It also will not invent a genuinely contrarian view unless the underlying research contains enough tension to support one.
Model Notes
Claude is strong at the intermediate “possible point of view” step because it handles nuance well.
GPT works well if you keep the stage outputs explicit and short. Model-agnostic overall as long as the staged flow is respected.
Related Resources
Browse PromptsSystem Prompt: Research Analyst
A system prompt for configuring an LLM as a structured research analyst that separates facts from interpretation, scores confidence, and flags gaps clearly.
Research & Analysis · Strategy & Planning
Prompt: Executive Summary from Dense Research
A summarisation prompt for compressing dense research into an executive brief without flattening the key caveats, risks, or implications.
Research & Analysis · Strategy & Planning
Skill: Market Research Synthesiser
A research synthesis skill for grouping messy inputs into patterns, tensions, implications, and next questions without losing uncertainty.
Research & Analysis · Strategy & Planning
Related Guides
Context Engineering > Prompt Engineering
Why the hard part is no longer phrasing clever prompts, but deciding what information the model should actually carry into the task.
Prompt Testing: How to Know If Your Prompt Is Good
A practical guide to prompt evaluation that goes beyond vibes and looks at repeatability, failure cases, and revision discipline.