Today, we’re announcing general availability (GA) for cognitive services, Azure ML and the AI workload in Power BI Premium. The AI workload will now be enabled by default for all P-nodes, as well as nodes A3/EM3 and above, allowing the users of those capacities to use cognitive services in dataflows right away. A new workload setting is added to separately enable Automated ML, which remains in public preview.
How customers are using AI in Power BI
Since text and image analytics were added in public preview in March, Power BI users have been able to use it on their own business data. The feedback we received from many customers about these features helped us to make great improvements!
One example of how text analytics is used in Power BI is from Kinepolis. Kinepolis is a cinema group operating 53 complexes throughout Belgium, The Netherlands, France, Spain, Luxembourg, Switzerland, and Poland, as well as 44 cinemas in Canada. The company uses Power BI AI capabilities to understand how they can improve their visitor experience by analyzing the sentiment of customer reviews and tweets.
“Kinepolis serves visitors in and from many countries, who vocalize their opinion through various channels. With text analytics in Power BI, we are able to structure and analyze all this feedback. It allows us to score the sentiment of our visitor reviews without investing in the data science behind it.” – Carlo Wambeke, technical lead BI at Kinepolis.
A new workload to support AI features
Power BI Premium allows you to dedicate capacity to the processes or projects that matter most, with the scale and performance you need. In addition to workloads for Paginated Reports and Dataflows, the dedicated workload for AI now gives the capacity admin more granular control over the capacity usage.
Capacity admins can set the maximum memory for the AI workload in the workload settings, under capacity setting in the admin portal. AI features need a minimum of 5GB. So, for example, the memory limit on an A3 capacity node needs to be at least 50%. You adjust the memory limit to prioritize other workloads in your premium capacity – and with the Premium Capacity Metrics app, capacity admins can track usage and performance of the AI workload.
The AI workload is used by cognitive services and Automated ML in Power BI. Automated ML can be enabled separately until it reaches general availability.
Ready-to-use text and image analytics
The GA of the AI workload makes it easier for users of a premium capacity to access cognitive services for text and image analytics. The first set of functions allows users to detect language of a text, score sentiment, extract key phrases, and identify topics depicted in images.
After the June release of the on-premises data gateway, cognitive services are now also supported with on-premises data sources.
Easy consumption of Azure ML models
With Azure ML integrated in Power Query Online, you can access all models created by you, or shared with you. Apply these models directly on your data in Power BI and repeat automatically on any new or updated rows during a data refresh.
Azure ML in Power BI is available to all users with Power BI Pro license.