What's New

AI Anomaly Detection scans a time-series metric for statistically unusual data points, so you can quickly tell whether a spike or dip is worth investigating before a meeting.

Click Analyze in the chart toolbar and select Detect anomaly. Holistics AI:

• Looks at the metric's full history
• Builds an expected range from past data
• Flags any data points that fall meaningfully outside it

From there, you can save the generated chart directly to a dashboard to keep monitoring that metric over time.

This is part of Smart Analysis — and it pairs with Analyze Changes: once you know a movement is statistically unusual, you can dig into exactly what drove it, without leaving your dashboard.

Analyze changes lets you right-click any data point that looks off and find out why it moved without filing a ticket or waiting on an analyst.

Every follow-up question beyond "what is the number?" — whether a change is real, what drove it, which segments caused it — currently requires analyst support. Analyze Changes brings that investigation directly into your dashboard.

Right-click any data point and select Analyze increase or Analyze decrease. Holistics AI runs a two-step investigation:

1. Checks whether it's worth investigating — using historical data to determine if the drop or spike is statistically unusual, or just normal variance you can safely ignore.
2. Explains what drove it — breaking down which dimensions, segments, or regions contributed to the movement, and by how much.

The result: business users can answer "why did this change?" on their own, in a few clicks, right where they're already working.

Admins can now configure MCP access separately from in-app AI access, per user or group.

Previously, one combined toggle granted a user both AI access inside Holistics and AI access via MCP together.

Now, granting AI access to a user or group lets you choose the scope:

• In-app only: use AI inside Holistics.
• In-app & MCP: also connect an AI client like Claude directly to Holistics via MCP.

This helps you grant MCP access only to users you trust to use an external AI client responsibly.

Learn more: Who Can Use AI

Field Filters can now be built directly from a metric, the same way you'd build one from a dimension or measure.

What's changed​

Previously, Field Filters only supported dimensions and measures as the source field - even though metrics are a first-class concept in the semantic layer alongside them. Now you can filter on a metric too, for example only showing rows where a computed metric like Churn Rate crosses a certain threshold.

Learn more: Field Filters

You can now redact data for AI.

Admins and Customer Success reps often need PII access to do their jobs - but that shouldn't mean AI gets it too. Previously, AI inherited the full permissions of whoever asked it a question, with no way to hold it to a stricter standard.

Now, you can configure your semantic layer so sensitive data stays visible to the people who need it and out of reach for AI, no matter who's asking.

Learn more​

• Redact data for AI

Custom currency is now live. If your team uses custom currencies or units, such as Bitcoin (₿), you can now set custom formats per field through the GUI input or directly in AML code.

Learn more​

• Data format
• AML number format

File history gives every file - dashboards, models, datasets - its own version timeline. See every version of a file with who changed it, when, and why, then restore just that file to any past version without reverting anyone else's work.

Previously, rolling back a past version worked at the project level: restoring reverted the entire project, including teammates' unrelated changes since then. Now you can scope a rollback to a single file. You can also duplicate a past version into a new file to compare side by side, or jump to the full project history to see everything that changed together.

Learn more: File History & Restore

Custom charts now live in AML. With this update, you can define them as code, create them through a GUI, and preview against real data:

• Define as code: Create a CustomChart block in a .chart.aml file. This lets you version-control your chart definitions in Git, validate their syntax, and protect them with dependency checks to prevent accidental deletion.
• Easy creation flow: Create a new chart in seconds with three simple starting points: chat with AI to generate it, pick a pre-made template, or build it from scratch.
• Preview while editing: As you write chart code, a preview panel renders the chart against real data. This lets you see and debug the output without leaving the editor.

⚠️ Legacy (non-as-code) custom charts in Admin Settings remain accessible for viewing, editing, and deletion, but new charts must be defined in AML. To migrate, use the Copy as code option in Admin settings > Custom charts.

Learn more: Custom charts | AML syntax reference

Color palettes can now be assigned at the theme level. With this update, you can:

• Apply color palette at theme level: Assign a palette to a theme so all charts on that dashboard use it automatically, eliminating the need for per-chart overrides. The palette also carries through to:
  • The theme thumbnails, so you can preview color palette before applying.
  • The explore modal, where charts inherit the dashboard theme's palette by default.
• Override at chart level: Individual charts can still set their own palette when needed.

Learn more: AML theme reference

Selecting a date range previously meant either navigating the calendar manually or knowing the exact phrase to type — there was no quick path for common choices.

The date picker now has a quick preset panel for one-click access to common ranges (Today, Yesterday, Last 7 days, Last 30 days, Month to date, Year to date). If you prefer typing, shorter shorthands now work too: ytd, mtd, and wtd. Existing long-form phrases continue to work — this is fully additive.

Learn more: Relative dates