Microsoft unveils AI boost for Power BI, new Fabric for data

AI and Power BI

Natural language processing (NLP) has long held the promise of enabling organizations to expand their use of analytics beyond just a small group of trained data scientists and analysts. But it never quite delivered on that potential.

While it simplified interacting with data by letting users model and explore data using natural language rather than code, the technology still required users to be data literate and phrase their queries and commands in a precise format.

At the heart of its limitation was a lack of vocabulary. And the result was that despite enabling business users to more easily interact with data, NLP did not substantially expand the number of users within organizations, which has remained stuck at about a quarter for well over a decade.

LLMs, however, have far more extensive vocabularies. As a result, when combined with analytics tools, they have the potential to eliminate -- or at least reduce -- the requisite data literacy to engage with analytics platforms and make analytics use more widespread.

It's pretty exciting to think about the potential of natural language to be the new interface between humans and computers. The potential to broaden analytical and BI capabilities to a truly broad audience is palpable.

In the six months since OpenAI released ChatGPT -- which marked a leap in generative AI and LLM capabilities -- Tableau, ThoughtSpot and Sisense are among the analytics vendors that have unveiled plans to infuse generative AI throughout their platforms. Power BI becomes the latest analytics platform to incorporate generative AI since the launch of ChatGPT, though all the capabilities combining analytics and generative AI -- irrespective of analytics vendor -- are in preview and not yet generally available.

No matter when Copilot in Power BI and the generative AI-infused tools from other vendors are ultimately released, they will be significant for analytics consumers, according to Doug Henschen, an analyst at Constellation Research.

"It's pretty exciting to think about the potential of natural language to be the new interface between humans and computers," he said. "The potential to broaden analytical and BI capabilities to a truly broad audience is palpable."

Once generally available, Copilot in Power BI will enable customers to use conversational language to ask questions of their data, create reports, generate and edit Data Analysis Expression (DAX) calculations, and create narrative summaries.

Henschen cautioned, however, that the audience for analytics won't broaden immediately upon the general availability of Copilot in Power BI and the generative AI capabilities of other vendors. Organizations will need to implement new security and governance measures to make sure their data is used safely and securely by its spate of new users.

In addition, he noted that while Microsoft's unveiling of Copilot in Power BI comes after some other vendors already unveiled integrations with generative AI tools, Power BI is not a follower in incorporating LLMs.

In 2021, Microsoft introduced a feature that uses GPT-3 to convert natural language into DAX, a coding language it developed to enable interactions with data. However, just like Copilot in Power BI and the generative AI-infused tools from other vendors, the natural language-to-DAX conversion feature remains in preview two years after its introduction.

"I would characterize [Copilot in Power BI] as just the latest LLM development related to Power BI," Henschen said. "Microsoft has been the pioneer in bringing generative AI to analytics and BI."

While Power BI, Tableau and ThoughtSpot are among the platforms integrating generative AI, security and accuracy concerns led longtime vendor SAS to delay such an integration.

SAS recently unveiled plans to invest $1 billion in AI over the next three years, but generative AI is not immediately part of those plans.

ChatGPT recently suffered a data breach, leading to worry about whether an organization's data might be at risk if such a breach were to happen again. And while mostly accurate, ChatGPT is not always accurate; if an organization were to base a decision on bad data, it could prove damaging.

So with Reggie Townsend, vice president of data ethics at SAS, serving on the U.S. Department of Commerce's National Artificial Intelligence Advisory Committee that works with President Joe Biden's administration, SAS is taking a cautious approach to generative AI.

Those vendors that are proceeding to integrate analytics tools and generative AI, however, cite the data governance capabilities of their platforms to ensure both the security and accuracy of data.

Microsoft is no different.

Microsoft uses plugins to connect its Copilots with other systems, and those plugins are governed by humans, according to Frank Shaw, Microsoft's chief communications officer.

"We've intentionally put humans at the center of our AI design, and our Copilots do not create content without a human first entering data that guides the AI response," he said on May 18 during a media briefing.

In addition to Copilot in Power BI, Microsoft unveiled Power BI Direct Lake and Power BI Desktop Developer Mode. Both are in preview.

Power BI Direct Lake is a storage mode with Power BI data sets that will eliminate the need for users to replicate data in order to unlock it within data lakes. Power BI Desktop Developer Mode uses a Git integration to enable developers to build data sets and reports.

A new Fabric, and more

Beyond its Power BI update, Microsoft revealed Microsoft Fabric in preview along with a host of new AI capabilities in Azure, the tech giant's public cloud computing platform.

Microsoft bills Fabric as a complete analytics platform that brings together data engineering, data integration, data warehousing, data science, real-time analytics and business intelligence.

The tech giant noted that many organizations piece together their analytics stack, incorporating BI capabilities from one vendor, AI and machine learning (ML) tools from another, and data ingestion and integration tools from still another. Even those organizations that use only Microsoft for their data management and analytics operations patch together different tools.

The results are complex systems and high costs.

Fabric is Microsoft's attempt to change that approach by providing users with all their analytics needs in one SaaS platform that both reduces complexity and helps manage costs. Fabric, once available, will combine seven core Microsoft analytics workloads, including Power BI, Data Factory, Synapse Data Engineering, Synapse Data Science, Synapse Data Warehousing, Synapse Real-Time Analytics and Data Activator.

In addition, the platform will be infused throughout with Azure OpenAI Service, providing Copilot capabilities that enable customers to use natural language to manage data pipelines, build AI and ML models, and analyze data.

"Today's world is awash with data, constantly streaming from the devices we use, the applications we build and the interactions we have," Shaw said. "Now, as we enter a new era defined by AI, this data is becoming even more important."

For organizations to derive value from the data, they need clean data managed by an integrated analytics system, he continued. Microsoft said it designed Fabric to be that integrated system.

As a result, Fabric could be the most substantial new data-related system Microsoft is unveiling this week, according to Henschen.

"[It's] the prospect of having a single platform for data with targeted, Azure OpenAI-powered interfaces -- including Power BI -- for many types of users," he said.

Beyond Fabric, among the more than 20 other new data management and analytics features in Azure are the following:

• Updates to Azure OpenAI Service in preview designed to better enable organizations to add their own data to their deployment and simplify integrating external sources.
• Azure AI Content Safety, a generally available service that helps enterprises create safer online communities with models designed to detect hate, self-harm, sexual and violent content in both images and text.
• A responsible AI dashboard in preview in Azure Machine Learning designed to help users evaluate large models built with unstructured data so that they can identify model errors, fairness issues and model explanations before the models are deployed.
• Model monitoring in preview so that users can track model performance, receive timely alerts and analyze issues.
• Updates to Azure Cosmos DB aimed at increasing developer productivity and optimizing costs.
• Updates to Azure SQL Database to enable users to build and quickly scale multiple applications.

"The common thread across it all is artificial intelligence," Shaw said, noting that Microsoft extended its partnership with OpenAI in January -- Microsoft is reportedly investing a total of $10 billion in OpenAI -- after initially investing in the generative AI vendor in 2019.

Looking ahead

The introduction of Microsoft Fabric recognizes that organizations use more than one vendor for their data management and analytics needs.

However, rather than attempt to seamlessly enable Power BI users to build stacks that include tools from other vendors, Fabric only brings Power BI together with Microsoft's data management and data science tools.

Henschen, therefore, said Microsoft should look to make it easier for Power BI users to create their own ecosystem for data management and analytics.

Most organizations store data on more than one cloud, and each cloud has its relative strengths and weaknesses.

"The Microsoft Fabric announcement acknowledges that companies have data in places in addition to Azure, including rival clouds," Henschen said. "My wish for Microsoft Power BI is that it would gain broader access to data than just data in Azure. Fabric seems like one route for such access, given that it's integrated with Power BI."

Another could be Direct Lake, he continued. But with the feature still in preview, it's not yet clear what kind of connectivity it will have beyond the Microsoft environment.

"I'd like to know if Direct Lake or other new capabilities will bring cross-lake data access directly to Power BI," Henschen said.

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.