# OneShot AI Mode

### Summary

**OneShot** is for when you have a specific question and want the **Generative BI AI agent** to do more work up front. It spends \~30 seconds (often less/more depending on scope) on **reasoning and validation** to return a polished package in a single shot: an **executive summary**, **headline metrics**, **supporting tables/charts**, and **suggested follow-up questions**. It’s ideal for updates to leadership, status readouts, or when you already know the outcome you want.

***

<figure><img src="/files/SaVfPoRFQyQpfInb9THX" alt=""><figcaption></figcaption></figure>

### How to get here (UI path)

* **Workspace → Channel → OneShot AI**
  1. Open your **Workspace**
  2. Choose the relevant **Channel** (bound to the right **Dataverse**)
  3. Click the **OneShot AI** tab/mode
  4. Enter a pointed question and press **Enter**

***

### What OneShot does (under the hood)

* **Frames the task:** interprets your question, aligns to your **Dataverse** (business lexicon, measures/dimensions, KPI definitions).
* **Validates assumptions:** checks time grain, filters, and metric formulas (e.g., *landed cost* = base + tariff + freight).
* **Runs a deeper pass:** aims to compute **scalars on full data** and compile **supporting artifacts** (tables/charts). If a visual would be heavy, it may render a **preview** first and offer “Promote to full data.”
* **Packages results:** returns an **executive summary** (what happened + why), **key drivers/outliers**, **supporting views**, and **suggested next questions**—ready to share or promote to **Data Studio/GenViz**.

***

### When to use OneShot (vs. Iterative)

* You **know your question** and want a “finished” answer in one go.
* You’re preparing a **leadership update** and need a crisp narrative with supporting views.
* You’re okay with the agent doing **more reasoning/validation** up front to save back-and-forth steps.
* You plan to **publish immediately** or **hand off to Data Studio** for production-grade visuals.

***

### What you can do here (at a glance)

* Get an **Executive Summary** that explains the result in plain language.
* See **headline metrics** (scalars calculated on full data whenever feasible).
* Review **supporting tables/charts** (preview or full, depending on complexity).
* Use **Suggested Follow-up Questions** to extend the analysis in one click.
* **Approve** the package and then **Publish / Pin / Share / Clone / Open in Data Studio (GenViz)**.

***

### Prompting patterns that work

* **Be specific:**\
  “Give me an executive summary of **1H 2025 revenue**: YoY change, top 5 drivers, and 3 risks to watch.”
* **Define outputs:**\
  “Return **3 bullet insights**, **2 KPI scalars** (Revenue, Gross Margin %), and **1 chart** (YoY by month).”
* **Ground definitions:**\
  “Use **landed cost** = base + tariff + freight; treat Tier-1 as ‘priority suppliers’.”
* **Constrain scope:**\
  “Focus on **US Retail**, **last 6 months**, exclude returns.”
* **Ask for checks:**\
  “List **assumptions** you applied and show the **logic/SQL** used.”
* **Request next steps:**\
  “Add **Suggested Questions** I should ask next and recommend a **Data Studio** visual.”

***

### Example walkthrough (executive readout)

1. **Ask:** “Provide a one-page executive summary of **Q2 revenue**: YoY growth, top 3 drivers, and 2 risks; include a small table of the top 10 SKUs by GM%.”
2. **OneShot returns:**
   * **Executive Summary** (narrative + bullets)
   * **KPI scalars** (Revenue, GM%, YoY%)
   * **Driver analysis** (categories/regions/vendors)
   * **Top 10 table** (preview or full)
   * **Suggested Questions** (e.g., “Drill into underperforming region”, “Show mix shift”)
3. **Review & Approve:** Check the **assumptions/logic** panel; ensure date ranges/filters are correct.
4. **Publish/Pin/Share** or **Open in Data Studio** to materialize the table/chart on **full data** and reuse the visual.

***

### Data scope & performance (how results are computed)

* **Scalars:** calculated on **full data** whenever feasible (authoritative numbers for comms).
* **Tables/charts:** OneShot attempts a **full run** when within guardrails; if heavy, it will **render a preview** and offer **Promote to full data** or **Open in Data Studio** to materialize at scale.
* **Clear indicators** show whether an artifact is **Preview** or **Full** so you know what you’re sharing.

***

### Quality checks before you **Approve**

* **Metric definitions** match your business lexicon (measures/dimensions, KPI formulas).
* **Filters/timeframe/currency** reflect the intended audience.
* **Preview vs Full:** promote visuals that must be production-grade; keep exploratory views in preview.
* **Narrative accuracy:** executive summary aligns with the evidence in the supporting views.

***

### Moving work forward

* **Approve → Publish:** Create a Message in the Channel.
* **Pin** key findings for visibility.
* **Share** a view-only link with stakeholders.
* **Clone** to tailor the message for different audiences.
* **Open in Data Studio / GenViz** to turn the outputs into reusable **cards, summary tables, and charts** that run on full data.

***

### Troubleshooting

* **Takes too long / times out:** Narrow scope (recent period, fewer dimensions) or request a **preview** visualization.
* **Narrative feels off:** Restate/lock definitions, specify the comparison basis (YoY vs QoQ), and ask the agent to list **assumptions**.
* **Missing artifact:** Explicitly ask for it (“Include a 10-row table with Supplier, Orders, Landed Cost, Tariff %”).
* **Heavy visuals needed:** Approve, then **Open in Data Studio** to materialize on full data.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.beye.ai/product-docs/oneshot-ai-mode.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.