Product Overview
Each mode is designed for a different decision context — balancing speed and depth so teams are not forced to choose between a fast answer and a rigorous one.
Not every business question needs the same level of investigation. A quick factual lookup is different from a root-cause diagnosis, and a root-cause diagnosis is different from a multi-angle strategic analysis. Carys modes let you match depth to the decision at hand — and will suggest the right mode if your question needs more than the current setting provides.
When to use: A single, direct business question with a clear factual answer.
What you get: The fastest path from question to answer. If your question would benefit from a more thorough mode, Carys detects the mismatch and suggests a switch — no need to restart.
When to use: A single KPI, issue, or trend that needs a rigorous, deep answer — not broad exploration.
What you get: Retains the full analytical engine: code execution, quality review agents, and advanced visualisation. Keeps scope tightly focused on your specific question and delivers deep precision without unnecessary breadth.
When to use: Strategic questions that need multiple angles, segments, and comparisons.
What you get: Full multi-step analysis with a reviewable plan. The Strategic Planner locks in consistent definitions before analysis begins, and a dedicated quality review agent independently validates findings at every step.
When to use: Advanced forecasting, modelling, or specialist analytical workloads.
What you get: Step-by-step collaborative analysis for high-complexity tasks where domain expertise and iterative refinement are essential. Designed for cases where the method itself needs to be developed alongside the findings.
Essential Insights, Deep Analytics, and Expert Analysis all benefit from architectural improvements introduced in v2.1 — making rigorous analysis more consistent, more reliable, and more transparent.
Before analysis begins, Carys investigates your data structure and establishes Query Guidelines — locked definitions for date ranges, how metrics are calculated, and what counts as each business term. Definitions like "churn" or "active customer" remain identical across every section of a report, preventing contradictions between pages.
A dedicated review agent independently evaluates the output of each analytical step. If it spots a discrepancy in calculations or logic, it intervenes to request corrections before findings are finalised. This runs up to three review rounds per analysis, significantly increasing numerical reliability without manual checking.
Cost estimates stream to your dashboard as the agent works, so teams can monitor resource usage dynamically throughout a long analysis — not wait for a summary at the end of a complex run.
If you ask a deep analytical question while in Quick Query mode, Carys detects the mismatch and recommends the optimal mode. Approve the suggestion and switch with a single click — the question carries over automatically.