The end of analysis is often a decision, not a dashboard. AI can help structure a concise explanation of a complicated result, provided it works from validated evidence and preserves uncertainty. The goal is to make a decision easier to assess, not to make the analysis sound more confident.
Start with the decision question, verified calculations, metric definitions, constraints and the person who will decide. Include plausible alternatives and the cost of being wrong. Do not ask the model to manufacture a recommendation from an unexplained spreadsheet.
What you will leave with
A one-page memo with a recommendation, alternatives, evidence, uncertainty and the next decision owner.
Step 1
Write the decision as a question
Name the actual choice and the person who owns it. Separate the decision from interesting background. For example, ‘Which campaign needs investigation first?’ is different from ‘Should we change next quarter’s budget?’ The second question needs additional evidence and authority.
The memo has a specific decision question and accountable reader.
Step 2
Create an evidence ledger
Give each material finding a source, calculation, period and caveat. Ask AI to summarize only the evidence provided and to flag gaps. Keep observations separate from interpretations. A reviewer should be able to trace every important number without searching through the original conversation.
Material findings have accessible sources and reproducible calculations.
Step 3
Compare options and uncertainty
Ask for realistic options, including maintaining the current approach when appropriate. For each, describe the expected mechanism, tradeoffs, evidence and what remains unknown. Avoid assigning invented probabilities or precise benefits. If the evidence does not support a choice, specify the additional investigation that would help.
Try this instruction
Write a one-page decision brief from this validated evidence. Lead with the decision and a supported recommendation. Compare alternatives, distinguish observations from hypotheses, include limitations and name what evidence could change the decision. Do not invent results, probabilities or authority.
Options include tradeoffs and do not invent certainty.
Step 4
Review for factual and decision quality
Use one pass to check numbers and citations, and another to ask whether the memo supports the actual decision. Remove background that does not affect the choice. Keep limitations next to the claims they qualify. Have the decision owner assess whether the available evidence is sufficient for the consequence of acting.
The brief is accurate, proportionate to the stakes and useful to the decision owner.
Step 5
Close the loop after the decision
Record the decision, rationale, owner and what will be observed next. Save the evidence version used at the time. When the result becomes visible, compare it with the original assumptions and update the workflow. This turns one AI-assisted memo into an accumulating record of how the team learns.
The decision and its follow-up evidence have a durable home.
When you get stuck
The memo is polished but the recommendation has no support.
Ask the writer to link the recommendation to specific findings and state what is still unknown. Withhold the recommendation when evidence is insufficient.
The team forgets why it made the decision.
Save the evidence, rationale and assumptions alongside the decision, with a named follow-up owner.
Terms explained
- Evidence ledger
- A compact record of findings, their sources, calculations and limitations.
- Decision owner
- The person accountable for choosing and authorizing the next action.
Sources and further reading
The workflow and templates are teaching guidance. The sources below support the underlying tools and concepts; product setup and availability can change.
Take it into your next task
Your completion checklist
0 / 5 checked
Checks are saved on this browser. Mark them only after you inspect the evidence. These are your checks, not an automated certification.