Building Dashboards, Reports & Scheduled Subscriptions

AI Insights 365 — On-Prem Edition gives BI builders a single place to design dashboards, freeze them into shareable reports, deliver them on a schedule, and push them downstream to Power BI. This reference walks through the full lifecycle: laying out a dashboard on the canvas, binding charts to live data, publishing and versioning a report, distributing it securely, and automating delivery. Each section follows the same flow you will use day to day, with explicit steps you can repeat.

How the pieces fit together

Before building, it helps to understand how the artifacts relate, because each one plays a distinct role and they are easy to conflate.

• A dashboard is the live, interactive design surface. It binds to datasources and reflects current data when refreshed.
• A report is a published, static snapshot derived from a dashboard. It freezes chart data at publish time and carries its own lifecycle and history.
• A subscription is an automated email delivery of a report on a cadence.
• A Power BI export/sync makes a dashboard available as a semantic model downstream.

In other words, you author once on the canvas, then publish, share, schedule, and export from that single source of design. Keeping these roles straight makes the rest of the workflow predictable: when you want interactivity and fresh data you work with the dashboard, and when you want a stable, distributable artifact you work with a report.

The dashboard canvas

Dashboards are authored on a drag-and-drop visual canvas. Rather than filling a fixed grid, you place and size components freely, arrange them across multiple pages, and preview real data as you build. The canvas is the design surface; everything else in this guide either consumes a dashboard or is derived from one.

A dashboard is made up of components you drag onto the canvas:

• KPI cards for single headline metrics.
• Charts drawn from 72+ interactive types.
• Tables for row-level detail.
• Shapes and layout elements for grouping, labeling, and visual structure.

Because the canvas is free-form and multi-page, you can model a single-screen executive summary or a paginated analytical workbook in the same project. Pages let you separate concerns — an overview page, a drill-down page, a data-quality page — without crowding one view.

Creating a dashboard

1. Open the Dashboards area and create a new dashboard.
2. Add a page, or keep the default first page.
3. Drag a component (KPI card, chart, table, or shape) from the component panel onto the canvas.
4. Position and resize it. Repeat to compose the layout.
5. Add more pages as needed for separate views.

At this stage components are present but empty. They become meaningful once bound to a datasource.

Binding components to a datasource

Every chart, KPI, and table draws from a datasource. AI Insights 365 ships with 24 connectors reaching SQL databases, Power BI, REST APIs, and File/URL sources, connecting directly to systems inside your network so data never has to leave it.

To bind a component:

1. Select the component on the canvas.
2. Choose the datasource it should read from.
3. Define the query or selection that returns the rows or aggregate the component needs.
4. Map fields to the component's roles — for a chart, that typically means category/axis, value/measure, and optional series or color.
5. Watch the real-time data preview render the result directly on the canvas.

The live preview is the core of the build loop. You are not designing against placeholder data and hoping it lines up later — you see actual values as you map fields, so layout and formatting decisions are made against the real shape of the data. If a category has far more members than you expected, or a measure spans an unexpected range, you find out immediately and adjust the query or the chart type on the spot rather than discovering the mismatch after the dashboard is shared.

Binding is also where you connect each component to the right source for its purpose. A single dashboard can mix datasources: a KPI card backed by a SQL query, a trend line from a REST API, and a table from a File/URL source can all sit on the same page, each reading from the connector best suited to it. Because every connector reaches its source directly inside your network, this remains true even when sources span different systems and networks.

Query result caching

To keep the canvas responsive and reduce load on source systems, query results are cached per datasource and per query. When several components share the same query against the same datasource, the result is served from cache rather than re-executed. The cache invalidates on refresh, so when you refresh the dashboard the underlying queries run again and the cache is repopulated with current data. In practice this means fast, repeated interaction during design while still giving you a clean, explicit way to pull fresh numbers when you need them.

Choosing the right chart

With 72+ interactive chart types available, the practical question is which family fits the message. The table below groups common types by analytical intent.

These are representative categories, not the full catalog — the same binding and preview workflow applies to every chart type. Start from the message you want the component to deliver, pick the matching family, then bind and format.

A few practical rules of thumb help here. Reach for a trend chart only when time is a meaningful axis; otherwise a comparison chart usually reads more clearly. Reserve composition charts such as pie and donut for a small number of parts that genuinely sum to a whole — they lose their advantage past a handful of slices. Use a gauge or KPI card when a single number matters more than its breakdown, and fall back to a table whenever the audience needs exact, lookup-style values rather than a visual impression. The breadth of 72+ types means you rarely have to compromise on fit; the discipline is choosing the simplest chart that carries the message.

Configuring charts and KPI cards

After binding, refine each component:

1. Set the title and any axis or legend labels.
2. For KPI cards, define the measure and, where relevant, a comparison or target so the single value has context.
3. For charts, confirm series, colors, and sort order.
4. Apply formatting (number formats, units) so values read cleanly.
5. Use the live preview to validate the result before moving on.

Publishing a report

A dashboard is a living, interactive design surface. A Report is a published, static artifact derived from a dashboard: publishing snapshots each chart's data at publish time. That snapshot is what makes a report stable and shareable — viewers see exactly the numbers that existed at publication, independent of later source changes or dashboard edits.

Reports move through defined states:

• Draft — a work-in-progress report not yet released.
• Published — a released report available to its audience.
• Archived — retired from active use but retained.

Revision history and rollback

Each report keeps a revision history. Every published revision is recorded, so you can review what changed across versions and roll back to a previous revision when needed. This gives report owners a safety net: if a new revision introduces a problem, the prior known-good version is one rollback away, and the history remains an auditable record of how the report evolved.

To publish a report:

1. Open the dashboard you want to publish.
2. Publish it as a report — this snapshots current chart data.
3. The report enters the draft/published lifecycle; set it to published when ready.
4. As you republish over time, each revision is captured in history.
5. If a revision needs to be undone, roll back to an earlier one.

Sharing reports securely

Reports are distributed using signed tokens, which keep distribution controlled even outside the application's authenticated users.

• Signed embed/share tokens — generate a token to embed a report in another application or share it via link. Because tokens are signed, they cannot be tampered with.
• Anonymous time-limited viewers — grant access to viewers without accounts for a limited window, useful for external stakeholders who need a look without onboarding.
• Per-report token revocation — revoke a token for a specific report at any time, immediately cutting off access granted through it.

To share a report:

1. Open the published report.
2. Generate a signed share or embed token for it.
3. For external access, configure an anonymous time-limited viewer link.
4. Distribute the link or embed the report where it is needed.
5. Revoke the token whenever access should end.

Combined with the platform's RBAC and AD/Entra SSO, token-based sharing lets you serve authenticated internal users and controlled external viewers from the same report without compromising on access control.

Scheduling email subscriptions

For audiences who should receive reports automatically, set up scheduled subscriptions that deliver by email on a cadence. Delivery uses your configured SMTP server, so mail flows through your own infrastructure.

Key options:

• Cadence — daily, weekly, or monthly delivery.
• Digest mode — bundle content into a digest rather than separate messages.
• Self-service or managed — members can self-subscribe, or admins can manage subscriptions on members' behalf.

To set up a subscription:

1. Ensure SMTP is configured for the environment.
2. Open the report and create a subscription.
3. Choose the cadence — daily, weekly, or monthly.
4. Enable digest mode if recipients should receive a consolidated message.
5. Confirm recipients: let members self-subscribe, or have an admin assign them.

Subscriptions turn reporting from a pull activity into a push one. Stakeholders who would otherwise have to remember to open a dashboard instead receive the latest report in their inbox on the schedule you set.

Digest mode is worth calling out for busy audiences. Rather than sending a separate message per report, a digest consolidates content into a single delivery, which keeps inboxes manageable when recipients follow several reports. The self-service option also reduces administrative overhead: members who care about a report can subscribe themselves, while admins retain the ability to manage subscriptions centrally for audiences that should not be opt-in. Match the cadence to how often the underlying data and decisions actually change — daily for operational reports, weekly or monthly for trend-oriented ones — so recipients get signal rather than noise.

Exporting and syncing to Power BI

AI Insights 365 dashboards can be exported and synced to Power BI as semantic models. This lets teams that have standardized on Power BI for downstream consumption keep their existing tooling while authoring in AI Insights 365. Exporting produces a semantic model representation of the dashboard; syncing keeps that downstream model aligned as the dashboard evolves.

This is particularly useful in mixed environments: analysts can build and iterate on the visual canvas here, then make the result available in Power BI for users and processes already anchored there — without rebuilding the model by hand.

Putting it together

The lifecycle is consistent end to end:

1. Compose a multi-page dashboard on the drag-and-drop canvas.
2. Bind KPI cards, charts, and tables to datasources, using the live preview and query caching as you design.
3. Publish a report to snapshot the data, and rely on states and revision history to manage it.
4. Share it with signed tokens or time-limited anonymous viewers, revoking access when appropriate.
5. Automate delivery with scheduled email subscriptions over SMTP.
6. Export or sync to Power BI where downstream consumers need it.

Because all of this runs on-premises — with the AI engine on Azure or air-gapped Ollama, 24 connectors reaching your in-network databases directly, RBAC, and AD/Entra SSO — your dashboards, reports, and subscriptions stay inside your environment from authoring through delivery.