Vibe Dashboards

Leapsome's vibe-coded custom view builder lets admins create live, AI-generated dashboards using plain-language prompts. Instead of configuring widgets field by field, you describe the visualization you want, and Leapsome generates it instantly from your live employee data.

Vibe-coded views support an expanded set of visualization types, including world maps, scatter plots, heatmaps, and treemaps, in addition to the standard chart types available in the regular dashboard builder.

Watch this video to see the Vibe Dashboards in action!

Table of contents

• When to use this feature
• What can a Vibe Dashboard show?
• Which data can a Vibe Dashboard access?
• What are the limitations of Vibe Dashboards?
• Who can see the data in a shared Vibe Dashboard?
• How to create a vibe-coded custom view
• How do I write a good prompt for Vibe Dashboards?
• When should I contact Leapsome Support?
• FAQ

When to use this feature

Use vibe-coded custom views when you need a visualization that is not available in the standard dashboard builder, for example:

• A world map showing headcount by country
• A scatter plot comparing attrition rates and average tenure
• A heatmap breaking down diversity metrics by department and location
• You would like to create a dashboard using an attachment; e.g., a PDF, or Excel file

This feature can be enabled via the setting ‘Analytics - AI vibe dashboard builder’ in the Feature Access Matrix by the super-admins:

What can a Vibe Dashboard show?

A vibe dashboard is built from the same data you can already reach in Leapsome with your own permissions. If you could look a number up in Analytics, a report, or a module overview page, a vibe dashboard can usually show it, and can combine several of these sources in one view.

Vibe dashboards work well for:

• Aggregates and breakdowns. Counts, averages, rates and distributions, cut by department, team, manager, location, legal entity, tenure or custom role.
• Cross-module views. For example, goal completion next to engagement scores per team, or absence patterns next to review completion.
• Visual formats the standard dashboard builder does not offer. World maps, scatter plots, heatmaps, treemaps, sankey and funnel charts, timelines, KPI scorecards.
• Operational lists. Overdue review tasks, open approval requests, learning path progress per person, upcoming absences.
• Trend views over a moving time window. Dashboards can use relative time ranges such as “last 12 months”, so they stay current every time they are opened.
• Playful formats built on workplace data. A name-learning game from colleague profiles, a team trivia board or a celebration wall for birthdays and anniversaries.

Every figure in a vibe dashboard is fetched live from Leapsome each time the dashboard is opened. The AI decides which data to use while you build the dashboard in the chat. After that, the dashboard only displays data. No AI runs when the dashboard is opened.

Which data can a Vibe Dashboard access?

There is no separate data catalog for vibe dashboards. The rule of thumb is:

If you can see it in Leapsome, a vibe dashboard can usually show it. If you cannot, it cannot.

In more detail:

• Your permissions apply. The AI works with the same access as the person building the dashboard. Data behind a permission or visibility setting your role does not have is not available, and there is no way to widen this from the chat.
• Your modules apply. Only modules your organization has enabled and set up can feed a dashboard. A module without data yields an empty dashboard section, not a magic one.
• Aggregated and list-level data is the sweet spot. Anything Leapsome already shows as an analytics view, a report, an overview table or a statistics page is a good candidate. The deeper the dashboard needs to go into individual entries, the more likely it hits one of the limits described in “What are the limitations of Vibe Dashboards?”
• The set of available data grows over time. We regularly extend what the AI can reach. If something you can see in Leapsome is not yet available in the chat, the AI tells you plainly and offers the closest alternative. Ask again after a few weeks, or check the release notes.
• You do not need to know the data model to build a dashboard. Describe the question you want answered in business terms and let the AI pick the data. The chat reply tells you which data was used in plain language, so you can check it matches your intent.

What are the limitations of Vibe Dashboards?

Knowing these limits before you start saves rebuilds.

• Data that is deliberately kept out. Everything in a vibe dashboard is visible to everyone who can open it, so the AI does not have access to data that is private to the builder as a person:
  • Goals whose visibility is set to “only me”
  • Private 1:1 meeting notes
  • The full detail view of an individual feedback set (feedback volumes and trends are available)
• Raw review answers in bulk. Review data is available as completion status, per-reviewer overall scores, and aggregated score analytics with the usual anonymity thresholds. A dashboard cannot pull every individual answer to every review question for all employees. If you need scores per person and per question, ask for review analytics broken down per question and per reviewee. Yes / No and free-text review questions are not scored and cannot be turned into numbers by the dashboard.
• AI analysis inside the dashboard. The AI helps you build the dashboard. It does not run inside it. A dashboard cannot summarize free-text comments, detect sentiment, classify answers into themes, or generate written insights when it is opened. What it can do with text is deterministic: show the comments, count them, filter them, or group them by an attribute that already exists in the data.
• Data from outside Leapsome. The dashboard runs in a sealed sandbox with no internet access. It cannot call external systems, load external images or fonts, or embed third-party content. An attachment you add in the chat helps the AI understand what you want to see (for example, a people strategy document or a mock-up), but the numbers shown always come from your Leapsome data.
• Bypassing permissions or anonymity. The dashboard cannot show data the builder cannot see in Leapsome, and survey and review anonymity thresholds apply exactly as they do in Analytics.
• Pixel-exact layouts. You can steer the design through the chat (colors, chart types, ordering, sections), and you can click an element on the canvas to make your instruction about that element. You cannot edit the generated code directly, and Leapsome does not hand-edit it for you.

Who can see the data in a shared Vibe Dashboard?

The dashboard uses the builder’s visibility. Data is fetched with the permissions of the person who built the dashboard, for every viewer. Someone you share the dashboard with sees what you can see, even if their own role is narrower. Choose the dashboard’s visibility groups with this in mind, especially for compensation, payroll or individual feedback data. The AI reminds you in the chat when a dashboard contains such data.

If the builder is deactivated or removed from Leapsome, the dashboard switches to each viewer’s own permissions. Data the viewer cannot access is reported as unavailable, and the rest of the dashboard still loads.

‘Show data’ drill-down

When a dashboard includes a record-level ‘Show data’ option behind its figures, that button is only shown to the builder and to super-admins. Other viewers see the same aggregates without the drill-down.

How to create a vibe-coded custom view

Step 1: Open the vibe-coded view builder

• Go to Analytics > Dashboards and Click ‘Create from vibe template' to use one of Leapsome templates (e.g. AI perception and adoption, Resilience Signals Deep Dive, Geographic distribution..)
•  Or click on 'New Dashboard' and select 'Vibe dashboard' to create from scratch
• Give your dashboard a name and set up your visibility.
• This opens a workspace split into a chat panel on one side and a live canvas on the other. The canvas displays a set of example prompts to help you get started.

Step 2: Describe your visualization

• Type your request in plain language in the chat panel. For example: “Show headcount by country on a world map,” or “Compare attrition rates and tenure as a scatter plot,” or something like “Break down age distribution by department as a heatmap”
• Leapsome interprets your request, selects the relevant data, and renders the visualization on the canvas. Each result shows which data source and segmentation was used, so the output is transparent.
• If your request is ambiguous or references data you do not have access to, Leapsome will ask a clarifying question instead of guessing.
• Add an attachment that might be helpful with the dashboard creation.

Step 3: Refine via chat

• You can continue refining the view through the chat. Each follow-up modifies the existing visualization rather than starting over. For example: “Now split it by department,” or “Change the time range to the last four quarters and switch to a bar chart,” or, if you upload an attachment, “I want to visualize progress on our people strategy as laid out in the attached doc from Notion. Please suggest a dashboard based on our data. Make it pretty!”

Step 4: That’s it!

• When you are happy with the result, you can exit the chat. Your vibe-coded dashboard is ready. The dashboard will appear in your Dashboards list alongside standard dashboards.

Viewing and editing saved vibe-coded dashboards

Once a vibe-coded dashboard is saved, it can be reopened at any time from the dashboards list. To continue refining it, click ‘Modify with AI’ to reopen the chat-and-canvas workspace and make further changes. Saved vibe-coded dashboards always reflect live data; they automatically refetch from the data source each time the dashboard is opened.

The following actions are also available within a saved vibe-coded dashboard:

• Click ‘Refresh data’ to manually reload the latest data.
• Click ‘Fullscreen’ to expand the dashboard to full screen.
• Click 'Start from scratch'. Please note, this removes your AI-generated dashboard and its chat history, then opens a fresh vibe dashboard builder. This action cannot be undone.

How do I write a good prompt for Vibe Dashboards?

• Say which question the dashboard answers and who will read it. “An overview for department heads of review completion and average manager scores per team for the current cycle” gives better results than “review dashboard”.
• Name the organizational cut you want: department, team, business unit, division, cost center or manager. These are different groupings in Leapsome and the AI will not mix them.
• Name the time range. Use relative ranges (“last 4 quarters”) so the dashboard stays current.
• Define fuzzy words. “Engagement”, “performance” and “health” mean nothing to the AI until you say which survey, review or goal data they should be based on.
• Keep custom scoring models simple and based on figures the dashboard can fetch. A model that depends on individual review answers or on interpreting free text cannot be built.
• Check the numbers against the corresponding Analytics view once. The dashboard uses the same data, so the figures should match. If they do not, tell the AI what differs.

Tips for follow-up messages

Follow-up messages are targeted edits. The AI changes only what you ask for and keeps the rest of the dashboard as it is. This works best when your message describes the result you want, not the problem you see. “This chart is wrong” gives the AI nothing to act on. “Show this as horizontal bars sorted from highest to lowest” does.

Describe the target state, not the complaint.

Point at the element you mean. Click the element on the canvas before you type. The AI then knows your instruction is about that chart, table or KPI, and you do not have to describe where it is. If you prefer typing, refer to the section by its visible heading (e.g., “the Review completion by team chart”).

One change per message, or a numbered list. Several unrelated changes in one paragraph get mixed up. Either send one instruction at a time, or number them: “1. Change the time range to the last four quarters. 2. Add a department filter at the top. 3. Rename ‘Health Score’ to ‘Engagement Score’.”

Be explicit about the dimensions the AI cannot guess:

• Chart type: bar, stacked bar, line, heatmap, table, KPI tile
• Sorting: by value descending, alphabetical, by date
• Time range: “the current review cycle”, “last 12 months”, “this calendar year”
• Organizational cut: department, team, business unit, manager
• Thresholds and colors: “mark anything below 70 percent red”
• Rounding and units: “one decimal”, “show as a percentage”, “scores on the 1 to 5 scale”

When a figure is wrong, give the reference. Say which value you expected, where you looked it up (the Analytics view, filters and period), and what the dashboard shows instead. The AI can then change the data scope rather than guess at a fix.

When you add data, name the area. “Add absence days per team from the absence data” works better than “add absences”. If the area is not available to you, the AI will say so instead of building something else.

Say what to keep. If you are changing one section of a dashboard you are happy with, add “keep everything else as it is”. This prevents the AI from touching sections you did not mention.

Undo instead of describing the mess. If a change made things worse, undo it from the chat and send a clearer instruction. Describing what went wrong after a bad edit usually leads to a second bad edit.

Do not restate the whole dashboard. Repeating the original prompt with a small tweak reads like a request to rebuild from scratch. Send only the change.

When should I contact Leapsome Support?

Contact Leapsome Support when the platform does not deliver something it should:

• The ‘Vibe dashboard’ widget is not visible even though the feature is enabled in the Feature Access Matrix (‘Analytics - AI vibe dashboard builder’) and the user has analytics permissions.
• A dashboard does not load at all, or shows an error for a data area every time it is opened, including after clicking ‘Refresh data’.
• A data area you can see in Analytics or in the module itself is reported as unavailable in the chat.
• Figures in the dashboard differ from the same figures in Analytics for the same filters and period.
• The AI keeps failing or timing out while building, independent of the prompt.

When you contact us, include the dashboard name, the prompt you used, the approximate time, and a screenshot of the message or the section that shows missing data.

We cannot help with:

• Editing or fixing the generated dashboard itself. The HTML and chart code of a specific dashboard is produced by the AI from your prompt, and the way to change it is a follow-up message in the chat. If a layout or chart is wrong, describe the change you want or select the element on the canvas and give the instruction there. If the result is still not right, start from a clearer prompt.
• Making data available that is not in Leapsome, or widening what the builder is allowed to see.
• Removing anonymity thresholds or the privacy exclusions described in “What are the limitations of Vibe Dashboards?”
• Building or validating custom scoring methodologies. The dashboard can display and combine the figures that Leapsome provides. Whether a custom index built on top of them is meaningful is a decision for your organization.
• Guaranteeing a specific visual design. The AI aims for a clean, consistent look based on your instructions, but exact pixel layouts are not supported.

FAQ

1. Does the dashboard update automatically?

Yes. Data is fetched every time the dashboard is opened, and ‘Refresh data’ reloads it on demand.

2. Can I share it with people who have fewer permissions than me?

Yes, and they will see what you can see. Set the dashboard’s visibility groups accordingly.

3. Why does the AI say a data area is out of reach?

Either the module is not enabled for your organization or your role does not include it. Your permissions in the dashboard builder are the same as in the rest of Leapsome.

4. Can I turn free-text survey comments into themes or sentiment?

Not inside a dashboard. The dashboard can list and filter comments. Analysis of free text happens in the Leapsome AI features built for that purpose, not in a vibe dashboard.

5. Can the dashboard show individual review answers per employee?

It can show per-person overall scores per review and aggregated scores per question through review analytics. It cannot pull the raw answers to every question for every employee.