What's New

You can also show everyone following AI Readiness where it stands, in the new AI Readiness Dashboard.

• Assess a semantic model or a data agent, preview an AI optimization, apply it and revert it, and generate, save and run a benchmark, all through the BI Pixie API at api.bipixie.com/v1. The API is available on every plan, the Free plan among them.
• Sign in with your own Microsoft account. There is no API key to create, store or rotate, and every call reaches what you reach in BI Pixie.
• Ask an AI assistant such as Claude to do the same work through the BI Pixie MCP Server on your own machine. Before a tool changes anything, BI Pixie shows you the change and waits for your decision.
• Share the BI Pixie AI Readiness Dashboard, a Power BI report that shows how ready your semantic models are, the optimizations you applied, benchmark accuracy over time, each A/B test and each data agent's score.

Data agent assessments are available in BI Pixie Cloud and in the BI Pixie Workload for Microsoft Fabric.

• Assess a data agent against Microsoft's published guidance and get one AI Readiness score from 0 to 100, with most of what it finds tracing to five rules Microsoft publishes, quoted inside the dimension each one explains. BI Pixie reads the definition read-only, involves no AI unless you select Ask your AI, and writes nothing until you approve a change.
• See when the semantic models are what hold a data agent down. Its score is limited by their AI Readiness, and that limit is a dimension of its own on the score card.
• Optimize a data agent's instructions, grounding, published description, and runtime, one change at a time. Your AI drafts the instructions, you edit them before they are applied, and BI Pixie publishes the data agent again after every change you approve.
• Survey your estate at no cost: the first assessment of each data agent is free on every plan, inside the same allowance of 500 free first assessments your semantic models sit in. The second assessment makes it a tracked item, and so does the first change you apply.

A/B tests are available in BI Pixie Cloud and in the BI Pixie Workload for Microsoft Fabric, on every plan.

• Compare a semantic model against the copy of it your team is optimizing, one semantic model answered through two different AIs such as a Fabric data agent and Copilot, or two data agents that differ only in their instructions.
• Both sides answer the same questions and are scored against the same correct answers, computed once per run from side A's semantic model, which is what makes the two results comparable.
• Before the run, BI Pixie reports how many of side A's correct answers side B's semantic model agrees with, so you can see where side B holds different data. A question whose value differs is still scored against side A's correct answer, and tagged Different data. The check asks no AI, uses none of your benchmark runs, and never stops the run.
• State your hypothesis before the run. One BI Pixie suggests comes back confirmed, rejected, or inconclusive; one in your own words comes back beside both scores for you to judge. Every question is paired side by side, with tiles for improvements, regressions, the questions where the sides differ, and whether both sides hold the same data.
• On the Free and Standard plans one A/B test run is included and does not draw on your benchmark runs. On the Pro and Enterprise plans, each side that runs uses one benchmark run. Side B's semantic model becomes a tracked item when the test runs.

AI Readiness is available in BI Pixie Cloud and in the BI Pixie Workload for Microsoft Fabric.

• Assess your semantic models across 12 AI Readiness dimensions, with the findings ranked by the fields your people actually query.
• Optimize your semantic models for AI with the help of your own AI and usage insights from the reports built on them. Nothing is written until you approve it.
• Create benchmarks to test the accuracy of Copilot in Power BI, a Fabric data agent, or a generic AI agent. A benchmark consists of business questions paired with correct answers computed from your own data, and you keep running it to detect regressions as your semantic models change.
• Survey your estate at no cost: the first assessment of each semantic model is free, for up to 500 semantic models on every plan.

Any Fabric tenant can add BI Pixie from the Workload Hub and track how Power BI reports are used, natively inside Fabric. This release includes the following new capabilities:

• Scalable telemetry processing with a Fabric notebook that builds your Lakehouse tables and semantic model on your own capacity.
• Bring Your Own Storage (BYOS): keep your raw telemetry in a storage account you control.
• Data residency: your events stay in your OneLake, in your capacity's region.
• AI assistants: query your engagement data in plain language with the Fabric Data Agent and the BI Pixie MCP Server, now generally available too.

Every Fabric tenant can add BI Pixie from the Workload Hub and start tracking how Power BI reports are used without leaving Fabric.

• Add BI Pixie to your tenant from the Workload Hub, then create BI Pixie items from the workspace New menu.
• The same dashboard, engagement metrics, surveys, and RLS auditing you know from BI Pixie Cloud, native in Fabric.

The BI Pixie MCP Server is now in public preview.

• You can run it locally in minutes with your own sign-in, or host it as an always-on endpoint for AI agents and teams.
• It is strictly read-only, with numbers that match your BI Pixie Dashboard exactly.
• It works with every BI Pixie deployment: Cloud, Self-Hosted, and the Fabric Workload.

A BI Pixie item lives right in your Fabric workspace, so you can track how your Power BI reports are used without leaving Fabric. It launched in private preview and is superseded by the public preview above.

• Create a BI Pixie item from the workspace New menu, governed by your Fabric workspace roles.
• The same dashboard, engagement metrics, surveys, and RLS auditing you know from BI Pixie Cloud.

Sign up, connect your Power BI workspaces, and start tracking user engagement in minutes. No setup or infrastructure is required.

• BI Pixie Cloud stays up to date, and new features and improvements are available automatically.
• Signing up is free, and the Free plan does not expire.
• Seamless onboarding: connect your Fabric workspace and add Pixies to your reports from the portal.

• Incremental refresh loads only new telemetry data on each run, cutting refresh time for high-volume environments.
• CI/CD deployment with zero Fabric item IDs to manage. One workspace name variable is all you need.
• A dedicated GitHub repository for customers on the Enterprise plan to deploy and customize the full pipeline.

• Lower latency and flexible architecture with no import refresh cycles.
• One governed data layer for telemetry, ready for scale.
• Native alignment with Fabric, RTI, and future workloads as Microsoft releases new capabilities like Fabric IQ and Ontology.

The new version groups insights by six key dimensions:

• Adoption: track who uses your reports and how usage grows over time.
• Engagement: measure session duration, interactions, and click patterns.
• Business Outcomes: quantify time savings, financial gains, and decision-making impact from embedded surveys.
• User Satisfaction: analyze feedback and NPS scores across reports and users.
• Performance: monitor report load times and identify slow pages and visuals.
• Security & Governance: audit Row-Level Security, data selections, and click-level activity.

BI Pixie exports reports from the workspaces you select, adds Pixies to them, and updates them back automatically. This works well with the Test stages of Power BI deployment pipelines.

Use the new Custom Visuals Heatmap to analyze certified, uncertified, and unknown custom visuals across your reports. Analyze visual clicks and performance by visual types and their state.

• Analyze data export events (Export to CSV, Excel) at the user, report, page, and visual level.
• Unprecedented lineage from data sources to business outcome. Compare the business outcome of different data stacks (e.g., Databricks vs. Snowflake).
• Assess business outcome by complexity score, including count of measures and total length of DAX expressions.
• Detection of scroll-down events in visuals.

• Paginated report views
• Export of data from visuals to Excel
• Report downloads as PDF and PPTX
• Analyze in Excel
• Email subscriptions

This requires the Self-Hosted deployment.

Detect ineffective design patterns and determine if excessive use of controls leads to low engagement.

The discussion about BI Pixie starts at 00:53:05.