AI PROJECT · DECISION BRIEF PROTOTYPE
Turn Fragmented Data into
a Decision-Ready Brief
A practical demonstration of how AI can synthesise information into options and risks, while an accountable owner confirms the brief before it's used to decide.
Project status
AI-Assisted Prototype
Business problem
Reports reaching executives are long and explain what happened, without stating what must be decided, between which options, and at what risk.
Why the problem matters
Executive time is finite. When the document is unclear, the decision is deferred — or made on instinct instead of evidence.
Who this is for
Built for executives and teams who want a concrete example of turning raw information into a decision brief with options, risks, and missing data made explicit, before considering how the approach might apply to their own work.
Project purpose
Create a one-page brief structure that names the decision, the options, the supporting evidence, the risks, and an auditable recommendation.
Chanikan's role
Designed the document structure, the prompts that force the model to state confidence levels and missing data, and the quality-review criteria.
Tools and technology
- A large language model for synthesis and drafting
- A self-designed brief template
- A recommendation quality checklist
- Fictional datasets for testing
Workflow overview
Workflow
- Gather raw inputs and name the decision to be made
- AI synthesises and drafts options with trade-offs
- AI states what data is missing and its confidence per conclusion
- A human reviews, edits, and owns the final recommendation
- The reasoning is recorded for later reference
Expected output
- A one-page brief with options and risks
- An explicit list of data still missing
- A written record of the decision rationale
Example application
Illustrative workflow example
Quarterly marketing budget reallocation decision
Input
- Sample performance and budget data
- Campaign result notes
- Market assumptions
AI-assisted process
- Draft options with trade-offs
- Flag missing data
- State confidence per option
Human Review
- Verify data assumptions
- Check the risk framing
- Confirm practical feasibility
Output
- A one-page brief ready for executive decision
Review by the accountable owner
The model must state confidence and missing data for every conclusion. A human reviewer confirms or corrects before the brief is used, and owns the final recommendation.
Responsible AI and data boundaries
- Examples use public, sample, or synthetic data
- Customer data, personal data, or confidential data should never be entered into a public demo
- AI output must be reviewed before it is used
- This project demonstrates an approach — it is not a fully autonomous enterprise system
Current limitations
- Tested against fictional scenarios only
- No automated fact-checking step yet
- Unsuitable for decisions requiring specialist data the model cannot access
Next development stage
Test the structure against a wider set of fictional scenarios and package it as a course template.
Chanikan's methodology
Chanikan Kanchanasalee
AI-Native Business Strategist specialising in Commercial Growth
Experienced in leading cross-functional work across Commercial Growth, Marketing, Sales, CRM, Data and Operations.