One AI Assistant Visualizations

One AI Assistant Visualizations lets you create charts, tables, and key values from your people data using natural language. Ask for things like how many people work here or hires and exits by department over the past 12 months with a forecast, and the Assistant automatically generates the best visualization type. Users with full access can refine results in several ways — by typing a short iterative prompt, editing the original prompt, or using the query panel to adjust filters, switch Series Types, customize formatting, or add features like forecasts, reference lines, and annotations.  When you're ready to share, you can pin a visualization to a Storyboard, save it to the Insights Library, or export it.

What you can do with Visualizations

The Assistant supports a wide range of prompt patterns, time selections, filters, and customizations. The following sections describe the main capabilities.

Generate charts, tables, and key values

The Assistant automatically selects a visualization type based on the shape of your data and the structure of your prompt. Supported types include bar, spline, line, area, area spline, scatterplot, geomap, table, and single value.

•       A metric by a dimension defaults to a bar chart

•       A time series defaults to a spline chart

•       A single metric by a geospatial dimension defaults to a geomap

Example: headcount by location as an area spline chart

Refine a visualization with iterative prompts

After the Assistant builds a visualization, you can type a short iterative prompt in the prompt bar to extend or refine the existing chart — without restating the full query. The Assistant merges the iterative with the current chart's metrics, dimensions, filters, and time selection, then re-renders the visualization.

Examples:

•       add terminations

•       split by department

•       show last 6 months

•       exclude contractors

•       change to a line chart

Display multiple metrics and dimensions

You can include several measures and breakdowns in a single prompt.

Example: hires, exits, and headcount by department and gender

Apply time selections

Use general terms to specify fixed or rolling periods or time models.

•       Past X months, quarters, or years can be requested, even if not available as a time model

•       Examples: July 2024, Q1 of 2023, past 17 days, last 6 months

•       Only the default calendar on your site is used

•       Year, Quarter, Month, and Day are the levels available; weeks are not currently supported

•       Past and Future time is not currently supported

Calendar selections work best with general time references like years, quarters, months, or specific dates. Custom labels such as FY 2024 or CY25 Q1 may sometimes work, but results are more consistent when you use standard terms like 2024 or Q1.

Filter results

Include or exclude dimensions or specific values directly in your prompt.

•       Examples: excluding contractors, only full-time employees, just finance and HR

•       You can also target specific dimension levels or nodes

•       Examples: Sup Org L3, Talent Acquisition

Sort, rank, and limit results

Ask for top or bottom results in plain language.

•       Use words like most, least, highest, and lowest

•       If no number is provided, five results are shown by default

•       Sorted charts default to horizontal bars

•       Examples: top 3 job families by terminations, top recruiters for time to fill last year

Drill Through into the data

Click on a chart element or table value to see the underlying records. Drill Through behaviour follows your site's configuration.

Add forecasts

Forecasts can be included directly in your prompt or configured in the Discover tab of the query panel.

•       Forecast periods default to four unless otherwise specified in the prompt or adjusted in the Discover tab

•       Additional forecast settings — such as forecaster type and confidence interval — can be edited in the Discover tab

•       Example: headcount over the last 12 months with a 3-month forecast

Customize every part of the chart or table

Use the Define, Design, Discover, and Describe tabs to modify metrics, dimensions, filters, Series Type, sorting, styling, and formatting. You can also add annotations or enable Embedded Insights such as lines of best fit and reference lines.

Export, pin, or save

Download or save your output for future use or sharing.

•       Export as an image or CSV

•       Pin to an existing Storyboard, create a new Storyboard, or save to the Insights Library

Provide feedback

Use the thumbs up or thumbs down buttons below the visualization to let the Assistant know whether the result was accurate. If you select thumbs down, you'll have the option to include a comment. Feedback is reviewed by the One Model team to help improve the Assistant.

How Visualizations works

When you submit a prompt, One AI Assistant uses OpenAI's large language model (LLM) to identify essential elements such as metrics, dimensions, dimension nodes, and time selections. For example:

•       Headcount is a metric

•       Cost Center is a dimension

•       Brisbane is a dimension node

•       Last 12 months is a time selection

These recognized elements are sent back to One Model, where One AI Assistant queries a vector database to find the closest matches in your data. Dimension nodes (for example, Engineering within Cost Center) are also available in this process, allowing for specific filtering requests such as “Show me headcount for the Engineering department only.”

With the matched metrics, dimensions, and time selections, One Model generates a query through its query engine. One AI Assistant then presents the data in the optimal or requested format — a table, chart, or key value. Using the query panel, you can see and modify the metrics, dimensions, time selections, and filters that were used to construct the visualization.

Create a visualization

You can access One AI Assistant from the prompt input bar at the top of any page in One Model. Creating a visualization is as simple as typing a prompt and reviewing the result.

•       Enter a clear, concise prompt in the prompt bar.

•       Click Go or press Enter.

•       Review the visualization the Assistant generates.

If configured for your site, you can also expand the dropdown to browse example prompts.

The Assistant uses generative AI to identify metrics, dimensions, dimension levels, time selections, filters, sorting, and forecasting instructions in your prompt and generates a visualization based on those inputs.

Example: Show me external hires and terminations for the past 8 quarters

One AI Assistant prompt bar

One AI Assistant example prompts

Use Visualization Mode

In AI Mode, the Assistant returns the most useful response for your prompt. Because results depend on your organization's configuration and permissions, it may not always produce a visualization when you expect one. To guarantee a chart, table, or key value, switch to Visualization Mode before entering your prompt. Every request then generates a visualization directly.

Interact with the data

Like Storyboards in One Model, results displayed in One AI Assistant are interactive.

•       Hover over a data point to reveal additional information

•       Click a series in the legend to hide or reveal it

•       Switch to a table view at any time

•       Click a data point to Drill Through to detail

Modify your prompt

You have two ways to adjust a visualization using natural language:

•       Type an iterative prompt in the prompt bar to extend the current chart — for example, add terminations or split by department. The Assistant merges your iterative with the existing chart parameters. See Refine a visualization with iterative prompts above for examples.

•       Edit or replace the original prompt text above the visualization, then click Go or press Enter. Use this when you want to start over with a different query.

Modify the query

One AI Assistant includes a full query builder that lets you adjust the result without rewriting your prompt, depending on your permissions. Users with both CanAccessOneAIAssistantVisualizations and CanDefineOneAIAssistantVisualizations can use the query panel to refine the result.

The Define tab shows how your prompt was interpreted and lets you refine it manually. You can update metrics, dimensions, Series Types, filters, and more.

Choose a visualization type

Select from Basic Chart, Table, or Key Value. Several Series Types fall under the Basic Chart option and can be defined for each metric.

Add or adjust metrics

•       Click the + button to add a metric, or the x to remove one.

•       Click a metric chip to open its settings.

•       Configure the metric using any of the following:

◦       Select a Series Type (spline, column, line, area, area spline)

◦       Toggle horizontal bar layout

◦       Enable Top/Bottom filtering and define the count

◦       Apply sorting (ascending or descending)

You can mix multiple metrics in the same chart or table.

Add or adjust dimensions

•      Click the + button to add dimensions, or the x to remove one.

•        Click a dimension chip to open its settings.

•        Use any of the following options:

◦       Pivot to X-axis

◦       Include or exclude specific levels or nodes

◦       Review selected values and clear them using Clear All

◦       Control dimension nodes

Dimension node selection icons indicate how each node is treated.

Pivoted dimensions appear on the X-axis in charts and as columns in tables. All other settings — selections, filters, and levels — work the same as standard dimensions.

Use time models

Time models are available for Time Period dimensions and enable dynamic selections that update automatically as time progresses. For example, if you select Today, the dimension always reflects the current date each time the site is processed.

Use pivot columns from your data model

Pivot columns let you include raw columns from your fact or dimension tables. They behave like pivoted dimensions in the output.

After you make your changes, click Run Query to regenerate the visualization with your updated selections.

Refine with the Design, Discover, and Describe tabs

In addition to the Define tab, three additional tabs let you refine and enhance your visualization.

Export and pin results

You can save and share your visualization directly from One AI Assistant.

•       Export Chart — Download a PNG image of the chart (Basic Charts only at this time)

•       Export Data — Download a CSV of the underlying data

•       Pin — Save to your Insights Library, a new Storyboard, or an existing Storyboard

Permissions

User access

User access to Visualizations requires the following:

•       One AI Assistant must be enabled for your company. See Permissioning One AI Assistant for setup details.

•       The Application Access Role CanAccessOneAIAssistantVisualizations controls user permission to create and view visualizations in Visualization Mode.

•       The Application Access Role CanDefineOneAIAssistantVisualizations separately controls manual editing in the query panel. Users with both roles can open the query panel and manually update the visualization.

•       Users with only CanDefineOneAIAssistantVisualizations do not have access to Visualization Mode.

Admin access

The Application Access Role CanConfigureOneAI controls configuration access. This permission allows administrators to manage all One AI Assistant setup and maintenance, including vector configuration. For more details on setup, access, and vector configuration, see the One AI Assistant Configuration Guide.

One AI Assistant relies on vector embeddings of configured metrics, dimensions, Storyboards, and Answers, stored in a secure vector database, to match user prompts to the right concepts and build relevant results.

Example prompt patterns

These example prompts show the kinds of inputs One AI Assistant can interpret. Replace the bracketed terms with your actual metric names, dimension names, and time selections.

Use these patterns to refine an existing visualization without rewriting the full query. 

Prompting best practices

Because the Assistant relies on a large language model and a vector database to interpret prompts and match them with your data, clear and deliberate prompts produce more accurate results. Keep the following best practices in mind.

Be specific, yet concise

Clearly state the metrics and dimensions you want to analyse. Avoid vague requests like show performance. Instead, ask for show headcount for the last 12 months by department. While specificity matters, avoid overloading your prompt with unnecessary detail — clear, concise prompts are easier for the LLM to interpret correctly.

Include dimension levels

If you know the specific level of a dimension you want, include it in your prompt — for example, terminations for the last 5 years by sup org L3. Otherwise, the Assistant defaults to level 1, and you need to manually adjust the level in the query panel.

Describe calendar selections generally

For calendar terms, general descriptions work better than customized formats. Use 2025, Q3 of 2024, Jan 2024, or 2020-01-12 instead of custom calendar labels like FY 2024 or CY25 Q1. Custom labels aren't searchable by the Assistant, so results can be inconsistent. One AI Assistant currently references only your site's default Time Period dimension.

Use familiar terms

While the vector database understands common synonyms, terms that match those in your data improve accuracy. For example, prompt with location instead of office site if that's how the field is labelled in your data.

Avoid ambiguity

If there are similarly named fields — like work location and home location — be specific in your prompt to avoid misinterpretation. Avoid terms that could apply to multiple fields when possible.

Include context

To narrow results, add context like time periods (last quarter, past year) and organizational units (finance team, sales department).

Refine as needed

If the result isn't what you expected, rephrase the prompt or add more detail. If you have Drill Through access, use it to inspect the metric definition and confirm the Assistant selected the correct one.

Use iterative prompts for small adjustments

When you want to extend or tweak a visualization rather than start over, keep your iterative prompt short and additive. Phrases like add, split by, exclude, show, and change to signal to the Assistant that you want to build on the current chart. For a brand-new question, type a full, standalone prompt instead.