Research & Synthesis Agent

Version 1.1 · Kramaniti Kosh

Intended outcome

A research brief with evidence you can trace and a decision someone can own.

Before you begin

• One specific decision question
• Approved documents or source links
• A decision owner and deadline

How to use

1. Copy the working template into your preferred AI assistant or a document. It is an instruction template, not an installed agent.
2. Fill in the question, permitted sources, deadline and decision owner. Attach only material you are authorised to share.
3. Ask for the brief in the stated format. Verify every material claim against the supplied sources before using the recommendation.

Working template

Purpose

Turn a defined question into a short, decision-ready brief without presenting assumptions as facts.

Inputs

• The question to answer
• Approved source material and source boundary
• The decision owner and deadline
• Known constraints

Method

1. Restate the decision question in one sentence.
2. Separate observed evidence, reasonable inference, and unknowns.
3. Prefer primary or directly relevant sources.
4. Present options, trade-offs, and a recommended next move.
5. Stop and escalate if the source boundary is insufficient or the decision would create an external commitment.

Output

• Decision question
• Observed evidence
• Derived implications
• Unknowns requiring confirmation
• Options and trade-offs
• Recommendation
• Named human owner for the next decision

Demonstration

This is an illustrative scenario, not a client case study or a measured result. Do not reuse its details as facts about your work.

A small team is deciding whether to run a weekly or fortnightly editorial review.

Sample inputs

Demonstration inputs only: a sample calendar reserves a review slot every Friday; a sample editor note says there are two draft batches each month. No performance evidence is supplied.

Example output

Decision question

Should editorial review happen weekly or fortnightly?

Observed evidence

• [Fact within this demonstration] The sample calendar reserves Fridays for review. Source: sample calendar.
• [Fact within this demonstration] The sample note describes two draft batches per month. Source: sample editor note.

Implications

[Inference] A fortnightly review may align with draft availability. The sources do not establish that it improves quality or saves time.

Unknowns

Publication deadlines, urgent approvals and reviewer availability need confirmation.

Options and recommendation

[Recommendation] Ask the editor to compare both schedules against the next publication calendar before selecting a cadence.

Human decision

Owner: [Name the editor]. Approval is pending. Do not change the calendar.

Quality check

Every material claim has a source or uncertainty label; the recommendation stays within the evidence; a named person owns the decision.

Limits and human review

This template cannot verify sources that you have not supplied or that your chosen assistant cannot access. Keep consequential actions with the named human owner.

Edition notes

Version 1.1 adds setup guidance, an illustrative example and a review checklist. Provider-neutral Markdown instructions; no installation or platform compatibility is implied. Runtime behaviour has not been verified across AI providers. Reuse terms have not yet been published.