Thuong

Orchestrator for open-ended analytics requests. Use when the user has a business question and a dataset but has not scoped which analysis type to run. Always runs data-quality-check → exploratory-data-analysis first, then dynamically selects any combination of diagnostic skills based on EDA findings, and closes with stakeholder-memo.

Use when a team makes a causal claim from observational data and the AI should assess counterfactual quality, alternative explanations, and next-step rigor.

Use when a user asks an ambiguous business, product, growth, operations, reporting, CSV, spreadsheet, dashboard, or data-analysis question and the AI should clarify goal, metric, dimension, grain, timeframe, baseline, filters, and decision context before answering.

Use when analyzing repeat behavior over time and the AI should clarify cohort logic, retention rule, cadence, and interpretation risks.

Use when reviewing a dashboard, KPI page, or reporting artifact and the AI should assess metric clarity, structure, comparability, and decision usefulness.

Use when a dataset, CSV, table, or export needs a structured reliability check before downstream interpretation.

Use when analysis requires new data pulls, modeling, instrumentation, or dashboard changes and the AI should create a precise request specification.

Use when reading out an A/B test or quasi-experiment and the AI should communicate effect, uncertainty, guardrails, limitations, and recommended action.

Use when a dataset needs a first-pass inventory of distributions, concentrations, time patterns, and candidate next questions.

Use when projecting a KPI or operating metric and the AI should produce a decision-ready forecast instead of a single-number guess.

Use when a user journey has sequential steps and the AI should identify where conversion breaks, how denominators change, and what segments matter.

Use when diagnosing a KPI change or planning an investigation and the AI should structure candidate explanations before diving into analysis.

Use when an analysis depends on a KPI, metric, ratio, column label, or business definition that could be interpreted in multiple valid ways across datasets, dashboards, or reports.

Use when investigating a KPI movement, spike, drop, or concentration in a dataset and the AI should structure evidence, candidate causes, confidence, and next checks in a disciplined way.

Use when grouping customers or users into actionable segments and the AI should define thresholds, profiles, and differentiated actions.

Use when an analysis of data, a CSV, a spreadsheet, a dashboard, or a KPI needs to be communicated to executives, PMs, operators, or cross-functional teams in a concise, decision-oriented format.