Decision Intelligence
How AskEdi handles decision-type questions with structured analysis, a mandatory recommended action, and a Decision Summary card.
Decision Intelligence is AskEdi's response mode for questions that ask what to do or where to focus. Instead of a descriptive answer, you get a structured analysis grounded in your data that drills into the strongest observed opportunity or risk and ends with a concrete recommended action.
Decision Intelligence responses are grounded in observed data patterns, not causal proof. The recommended action is a data-grounded starting point for investigation, not a directive to execute immediately. Use the validation metric in the Decision Summary card to confirm the direction before committing resources.
When Decision Intelligence Activates
AskEdi classifies the intent of your question automatically and routes it to the appropriate response type. Questions that typically activate Decision Intelligence include:
- "Where should I focus to improve revenue?"
- "Which product category should I prioritise?"
- "What is driving churn and what should I do about it?"
- "How can I reduce operational costs?"
- "Which region has the most untapped opportunity?"
You do not need to use a special keyword or command.
What a Decision Intelligence Response Includes
A Decision Intelligence response is a structured analysis followed by a Decision Summary card. The depth of the analysis depends on the complexity of the question.
A deep response covers 4 to 6 sections, each advancing or qualifying a central thesis. Typical sections cover the trend over time, the primary driver or dimension, a composition breakdown, a risk or anomaly, and a recommended action.
At least one section is always titled with an action-oriented phrase, for example "Recommended action", "Priority action", or "Next best move".
Each section is grounded in actual query results from your data. AskEdi performs at least one diagnostic drill-down into the strongest observed opportunity, risk, or performance gap before recommending action.
A simple response is 3 to 4 sentences covering the primary takeaway, the key evidence, and a recommended action. It uses the same observational language and data grounding as a deep response, but without the multi-section structure.
Decision Summary Card
The Decision Summary card appears at the end of every Decision Intelligence response. It contains five elements:
| Element | What it contains |
|---|---|
| Recommended action | A concrete directive using imperative language: Prioritise, Reduce, Reallocate, Investigate, Protect, Test, or Pause. Always supported by observed evidence from your data. |
| Evidence | The specific data observation that grounds the recommendation. A query result, concentration, trend, or gap from your source. |
| Scenario-based upside or risk | What the observed opportunity or gap suggests could happen if action is taken or not taken. Stated as a scenario, not a prediction or guaranteed outcome. |
| Confidence | High, medium, or low. The strength of the evidence, always accompanied by the observable reason: concentration percentage, consistency across time, sample size, or dominance versus the next-ranked dimension. |
| Validation metric and time window | The metric to watch and the period over which to confirm the action is working. Use this before committing resources: it tells you what to measure to verify the recommendation is directionally correct. |
If the impact cannot be quantified from the available data, AskEdi states the action, evidence, confidence, and validation metric without inventing a figure. A response that says "impact cannot be quantified from current fields" is being accurate, not evasive.
What This Looks Like
A shortened example, for a question like "where should I focus to improve revenue?":
Revenue by Region The West region contributes 41% of total revenue, the largest share of any region, but its growth rate over the last two quarters is flat compared to a 12% rise in the East region.
Recommended Action Prioritise the East region's growth pattern for deeper investigation before reallocating budget, since it is the strongest observed opportunity in the current data.
Decision Summary (highlighted card) Recommended action: Investigate what is driving East region growth and evaluate whether it can be replicated in West. Evidence: East grew 12% over two quarters versus flat growth in West, which holds 41% of total revenue. Scenario-based upside: if the East pattern holds, applying it to West's larger base could meaningfully increase total revenue, though this is a scenario, not a guarantee. Confidence: medium, because the growth gap is clear but the underlying driver has not yet been isolated. Validation metric: track West region revenue growth rate over the next two quarters against the East benchmark.
The real response is longer and grounded in your own column names and numbers. This shows the shape, not the content.
Observational Language Guarantee
AskEdi uses observational language throughout Decision Intelligence responses. The rules that govern this are applied consistently.
- Identify the strongest observed concentration, trend, or gap in your data
- Drill down into the primary driver before recommending action
- Quantify the scenario range where the data supports it
- State confidence with the observable reason
- Qualify recommendations when evidence is weaker: "based on the observed concentration" or "given the current trend"
- Provide a validation metric so you can verify the recommendation before acting
- Use causal-certainty language ("will cause", "guarantees", "proves", "will yield") unless your data contains experimental results, intervention history, or explicit causal fields
- Collapse an observed driver into a root cause when causal fields are absent
- Recommend actions that require systems or data not present in your source
- Hand the decision back with vague phrases like "use this to decide"
When evidence is low confidence, that is stated explicitly with the reason.
How Decision Intelligence Combines With Other Response Types
Decision Intelligence is not a separate response type competing with Root Cause Analysis, Forecasting, What-If Analysis, and Category Comparison. It is a layer of action-oriented framing that AskEdi applies on top of whichever of those response types actually answers your question, whenever your phrasing asks for guidance on what to do.
A Root Cause Analysis response already closes with a Decision & Impact Summary card, and a What-If Analysis response already closes with an Operational Recommendation & Assumptions card. Asking "why did sales drop and what should we do about it" produces the same Root Cause Analysis structure you would get from asking "why did sales drop" alone, decision framing does not add a second summary card on top.
A plain Forecasting response closes with forward-looking monitoring guidance, what to watch for, not a directive action. Asking "what should we do about the sales forecast" replaces that closing section with a full Decision Summary card instead.
For a narrative question with no trend, forecast, simulation, or category test behind it, asking for direction ("where should I focus to improve revenue") adds one extra body section naming the priority area, followed by the Decision Summary card described above.
Frequently Asked Questions
Next Steps
Root Cause Analysis
How AskEdi answers why a metric moved, with a real regression, contribution ranking, and correlation test.
What-If Analysis
How AskEdi models a hypothetical change instantly, using a sensitivity measured from your own data.
Methodology Notes
Inspect the exact statistical method and test result behind a response, written by the system, not the AI.
AskEdi
Back to the AskEdi overview: how a chat works, response types, and how to get started.
Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.
Last updated on