> For the complete documentation index, see [llms.txt](https://docs.ins-pi.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ins-pi.com/command/features/command-dialog.md).

# Command Dialog (AI assistant)

Command Dialog is the assistant built into Command; it answers portfolio questions with graded evidence, runs guided analyses, builds content for you and writes only after you confirm.

**Command Dialog** is the assistant built into YouDesign Command. You type a question or a request in plain words, and it answers from the records in your ServiceNow instance, shows how much the answer can be trusted, and opens the content it builds in a tab. It runs in your browser and on your ServiceNow instance, and it does not send your questions or your data to ins-pi or to any third party.

## What it is

A panel on the right of the screen. Its header reads **ins-pi Transformation OS™**, with **Reset thread** and **Close panel** buttons. Below the conversation is the message box, *Let's transform together*. You can also ask from the **Command Transformation** box on [Home](https://docs.ins-pi.com/command/the-portal/home.md).

<figure><img src="https://414746897-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fx4mE784snR9Eb8oVrk4e%2Fuploads%2Fgit-blob-a53eceae689e9e11c0362e3f8328595fd1f58c72%2Fcommand-copilot-decommission.png?alt=media" alt="The Command Dialog open on the right with an answer and its chips, and the attached-content chip above the message box."><figcaption><p>The Command Dialog with an answer and the attached-content chip</p></figcaption></figure>

What you can ask today:

| You want to                       | Try                                                                                    | What happens                                                                                                                                                          |
| --------------------------------- | -------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| See what it can do                | `what can you do?`                                                                     | Where to start, with **Show everything I can answer** for the full list.                                                                                              |
| Learn a term                      | `what is a SWOT?`, `measure vs score`, `what is the TOGAF ADM?`                        | A short explanation. Architecture terms such as ArchiMate, Zachman, baseline and target, and TCO, OPEX and CAPEX are explained without starting a run.                |
| Run a guided analysis             | `which applications should we rationalise?`                                            | A [guided use-case](https://docs.ins-pi.com/command/features/guided-use-cases.md) runs step by step in the conversation.                                              |
| Find retirement candidates        | `which applications should we decommission?`                                           | A ranked shortlist, with chips to show it as a TIME Analysis, on a Roadmap or as a list, or to open a recommendation on one of them.                                  |
| Decide about one record           | `should we retire PayGate?`, `shall we keep AssetMan?`                                 | A guided run scoped to that record, ending in an [Advisory Report](#the-advisory-report).                                                                             |
| Rank by cost                      | `what are our most expensive applications?`                                            | A ranked table with its total, saying which cost it ranks by, for which period and in which currency. See [Rank applications by cost](#rank-applications-by-cost).    |
| Get a brief                       | `how much do we spend on IT?`, `brief me`, `what are my top 3 decisions this quarter?` | A short answer that leads with the figure. See [Get an executive brief](#get-an-executive-brief).                                                                     |
| Ask about your estate             | `how many incidents last month?`, `which CIs have no owner?`                           | A count from the full table, with the table total beside it, for example *0 last month (70 in total)*.                                                                |
| Check compliance                  | `are we compliant with DORA?`, `what are our GDPR gaps?`                               | Coverage from the [Framework Intelligence Center](https://docs.ins-pi.com/command/features/framework-intelligence-center.md), with what is missing. Not legal advice. |
| Plan a cost reduction             | `how can we reduce cost by 25% by end of 2027?`                                        | A guided [cost reduction assessment](#plan-a-cost-reduction).                                                                                                         |
| Build a dashboard                 | `create a CFO dashboard`                                                               | A preview, then **Yes, create it**.                                                                                                                                   |
| Create content                    | `create a roadmap`, `create a relationship editor`                                     | The content opens right away. Add `and share it with Klaus` and the share is offered as a chip.                                                                       |
| Show it another way               | `show them as a roadmap`, with an R-Analysis open: `show invest and migrate`           | The records of the open content, as another content type.                                                                                                             |
| Work with what you just made      | `rename it`, `share it with Klaus`, `export it as PDF`                                 | The action, on the content you just created. Sharing still asks you to confirm.                                                                                       |
| Open a record                     | `open PayGate`                                                                         | The record opens as a Form.                                                                                                                                           |
| Find who owns a record            | `who owns PayGate?`                                                                    | The record's owners, from its owner fields.                                                                                                                           |
| Understand what is on screen      | `explain this roadmap`, `what are we looking at?`                                      | A summary of the open content.                                                                                                                                        |
| Change a setting (administrators) | `add a phase Retired to the internal lifecycle for servers`                            | An [editable proposal](#change-settings-administrators) to confirm.                                                                                                   |
| Take the tour                     | `give me a demo`                                                                       | A guided walkthrough.                                                                                                                                                 |

## When to use it

* You have a portfolio question and want an answer with its evidence, not a spreadsheet.
* You want a starting point built for you, such as a dashboard or a shortlist, and then refine it by hand.
* You want to understand or explain the content you have open.

## How to

### Open it and ask

1. Click **Command Dialog** on the right edge of the screen. On a phone, open it from its button in the top bar. The message box is ready to type in.
2. Read the greeting, for example "Good morning, Alex. What would you like to work on?". If you left an analysis unfinished, it offers to pick it up: see [Pick up where you left off](#pick-up-where-you-left-off).
3. Type your question and press **Enter**. Use **Shift+Enter** for a new line.
4. While it works, a short status such as **Working…** or **Checking the evidence…** shows. Click the stop button in the message box, or press **Escape**, to stop.
5. Click a chip under the answer to take the next step. Follow-up questions are answered against the last result.

To start over, click **Reset thread**, or type `/clear`. When the assistant is idle, **Escape** closes the panel.

### Ask about what is on screen

The chip above the message box shows the content the assistant will use as context: the icon and title of the item in your active tab.

* It **follows the active tab**. Switch tabs and the chip switches with you.
* Opening or creating an item attaches it. Rows you select in a table are attached too, shown as *N selected* with **Clear**.
* Home and full-page views such as Governance or the Glossary attach nothing.
* Click **×** (**Remove attached content**) to ask without context. This applies to the current tab only.

Then ask, for example, `explain this roadmap` or `what are we looking at?`.

### Pick up where you left off

A guided analysis is saved once its scope is confirmed. When you come back, the greeting offers:

* **Resume** , with how far it got, such as *Guided run, result ready*;
* **Show open analyses**, which lists them with **Discard** for those you no longer need;
* **Continue:** for proposals waiting for approval.

You can also type `continue`, or `show my open analyses`.

### Confirm a write

Command Dialog writes to your instance only when you confirm.

1. Ask for a dashboard.
2. Read the preview. It says what will be created and states that nothing is written yet. It plans only widgets whose source has data.
3. Click **Yes, create it**, or type `yes`. Click **Cancel** to drop it. The card then shows what you answered.

<figure><img src="https://414746897-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fx4mE784snR9Eb8oVrk4e%2Fuploads%2Fgit-blob-c528658844cae11204258f9dbf6cf48c5e3b6595%2Fcommand-copilot-dashboard-preview.png?alt=media" alt="The Command Dialog showing a dashboard preview that says what would be created, with the Yes, create it and Cancel chips."><figcaption><p>A preview waits for Yes, create it</p></figcaption></figure>

Content you ask for directly, such as a roadmap or a SWOT, opens right away, the same as using the create menu. Command Dialog does not delete content in bulk.

### Rank applications by cost

1. Ask `what are our most expensive applications?`.
2. Read the table: it ranks by a cost measure when one has data, then by a cost score or a cost field. It says what it used, the period and the currency, and shows the **Total:**.
3. In the table that opens in the main window, use **Rank by** to switch the measure, and the slider to choose which rows show. The table changes in place.
4. Under the answer, pick from **Choose a view** to open the same applications as a TIME Analysis, R-Analysis, Static List, Roadmap or Scenario.

When there is no cost data, Command Dialog says so instead of ranking.

### Get an executive brief

Ask the way an executive would, for example `how much do we spend on IT?`, `where can we save 10%?` or `what should I focus on?`. The answer is short: a headline, up to three facts, a caveat and a **Next step:**, followed by chips.

* Click **Show the full analysis**, or type `show the full analysis`, for the complete answer.
* Type `so what?`, `bottom line` or `make it board-ready` to have the last answer restated briefly.
* Name a role, such as `the CIO view`, to open that role's dashboard.

### Change settings (administrators)

A Command administrator can change configuration by asking: lifecycles, measures, scores, data-quality checks, approval policies, personas, governance bodies, assessment rules, landing pages and General settings. For example:

* `create a measure Energy Cost in EUR`
* `check that owner and cost center are filled on business applications`
* `prepare a landing page for Jane Smith from the dashboard CFO Overview`

1. Read the proposal card. Nothing is saved yet.
2. Edit its fields and rows: reorder, add, or remove the rows you added.
3. Click the confirm button, such as **Create lifecycle** or **Save setting**, or **Cancel**.

Existing rows are never removed this way. Anyone else who asks is told the change needs the Command Admin role, and nothing is changed.

### Plan a cost reduction

Ask, for example, `how can we reduce cost by 25% by end of 2027?`. A **Cost reduction assessment** opens in the conversation and asks one question at a time. Anything you already said, such as the percentage or the deadline, is filled in.

1. **Target**: the percentage reduction. Confirm with **Use this target**.
2. **Deadline**: taken from your request when you gave one, or pick a date.
3. **Measures**: the cost measures maintained in your instance.
4. **Areas**: **Applications**, **Contracts**, **People** or **Infrastructure**.
5. **Scope**: which records to cover. **Assess every record**, or filter them. Command counts them and shows every filter in words.
6. **Basis**: **Recurring cost at the deadline** or **Savings realized during the period**.
7. **Baseline**: **Approved actual period**, **Approved budget** or **Approved forecast**.
8. **Period**: **Last 12 months**, **Last calendar year**, **Last financial year** or **Current financial year**.
9. **Constraints**: for example **Maintain service commitments** or **Exclude workforce changes**.
10. **Data readiness**: how complete the cost data is. **Proceed**, **Proceed with partial data**, or fix the data first.

The result shows the baseline cost, how many records have a cost, the target saving, the gap, and where the cost sits. A partial result says so. The candidate levers show where cost sits, not savings that have been verified. Click a chip's **×** to change an answer, or **Clear all** to start again.

## Trust and evidence

Every recommendation says how much it can be trusted and why.

| On screen           | Meaning                                                                                               |
| ------------------- | ----------------------------------------------------------------------------------------------------- |
| **Ready to decide** | The evidence is strong enough to act on.                                                              |
| **Needs approval**  | The call should be made by a named person.                                                            |
| **Not ready**       | The evidence is too thin. The report names the first gap to close.                                    |
| Confidence **/100** | How certain the answer is. If the evidence holds it back, the report says so and what would raise it. |

Each piece of evidence carries a grade:

| Grade           | What it means                      |
| --------------- | ---------------------------------- |
| **attested**    | A person signed off on this value. |
| **measured**    | Computed from governed data.       |
| **declared**    | Somebody typed it into a field.    |
| **unsupported** | Nothing backs this read.           |

Answers that contain a table open the table in a tab, and the conversation points to it.

If you challenge a figure, Command Dialog explains what it rests on. It corrects a number only when its source is wrong, never because it is disputed, and asking the same thing twice gets the same answer. It does not rate or blame people, and it declines requests such as granting access in bulk or exporting people's details.

## The Advisory Report

An **Advisory Report** turns an analysis into a decision document. It opens in a tab and re-reads your data every time you open it.

* **From a guided analysis:** click **Create advisory report** when a run has its results. The report says whether the decision is ready, what it answers, the result, how the data was read, what to fix to make it decision-ready, the records in scope, and the decisions taken. **Continue the run in Command** picks the analysis up again.
* **For one record:** the recommendation report shows the verdict and **Decision on the record**, **Why: the factors behind it**, **Data quality behind this call**, and **Decisions on** with **Notify owner**.

<figure><img src="https://414746897-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fx4mE784snR9Eb8oVrk4e%2Fuploads%2Fgit-blob-d2eba544be81ead7b63c30cc37988273da1c3cee%2Fcommand-copilot-advisory-report.png?alt=media" alt="An Advisory Report with the verdict stamp, the confidence meter and the factors behind it."><figcaption><p>An Advisory Report opened from the Command Dialog</p></figcaption></figure>

1. Read the verdict and the confidence.
2. Work through what it asks you to fix, or use **Notify owner** to ask the owner or steward to fix the data. You see who will be told before anything is sent.
3. Click **Re-run after fixes** once the data is fixed.
4. Record the decision under **Decisions on** . **Decision on the record** then says whether it matches the recommendation.
5. Click **Download PDF** to share it outside Command.

When the record's data quality says **Do not rely**, the report holds the verdict at **Not ready** and says the data must be reviewed first, whatever the other factors say.

Reports made before guided analyses existed open read-only, with **Re-run as a Guided use-case** to refresh them.

{% hint style="success" %}
An Advisory Report is advice. Recording a decision does not change any lifecycle, investment or record. A named owner has to act on it.
{% endhint %}

## Privacy

| Question                                        | Answer                                                                                                                                                                                                                     |
| ----------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Where is my question processed?                 | In your browser and on your ServiceNow instance.                                                                                                                                                                           |
| Does anything go to ins-pi or to a third party? | No. Your questions and your data are not sent to ins-pi or to any third party.                                                                                                                                             |
| Where does the data come from?                  | Your ServiceNow instance. Command Dialog reads records with your own permissions and writes content and confirmed changes back to the same instance.                                                                       |
| Does the tour change my data?                   | No. The guided demo builds its examples on your device and changes nothing in ServiceNow.                                                                                                                                  |
| What about other AI assistants?                 | Connecting Claude or another AI assistant to Command is a separate product, the MCP Connector. It is off unless your administrator switches it on, and using it sends the data you can see to the AI provider you connect. |

Command shows these statements to your administrator on the **License** page. See [Where your data goes](https://docs.ins-pi.com/command/administration/license.md#where-your-data-goes).

## Good to know

* If you ask for an analysis that has no method yet, Command Dialog says so instead of guessing, and offers **Record this as an idea** to send it to the [Ideation Portal](https://docs.ins-pi.com/command/features/ideation-portal.md).
* A shortlist says how many records qualify and how many are shown, for example "Eliminate 11 (top 10 shown)".
* Content Command Dialog creates appears in [My Space](https://docs.ins-pi.com/command/the-portal/my-space.md) under its own name, ready to use.
* On a phone the panel covers the screen, and it stays open when an answer opens a table or creates content.
* An administrator can switch the assistant off for everyone in **Global Settings > General** (**Command SLM assistant**). Advisory Reports already created stay available.

## Related

* [Guided use-cases](https://docs.ins-pi.com/command/features/guided-use-cases.md)
* [Scoring](https://docs.ins-pi.com/command/features/scoring.md)
* [Data Governance](https://docs.ins-pi.com/command/features/data-governance.md)
* [Measures](https://docs.ins-pi.com/command/features/measures.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

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

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.ins-pi.com/command/features/command-dialog.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.
