AI assistant from Tableau targets efficiency, deep analysis
The analytics vendor's copilot can recommend questions that might lead to otherwise undiscovered insights as well as understand follow-up questions that result in deep analysis.
Tableau on Tuesday unveiled the beta testing of Einstein Copilot in Tableau, a new AI assistant designed to make users more efficient as well fuel deeper data analysis. General availability is scheduled for this summer.
Based in Seattle, Tableau is a longtime analytics vendor whose platform aims to enable both data experts and self-service business users to explore and analyze data to inform decisions.
The vendor was acquired by CRM giant Salesforce in 2019. Since then, Salesforce and Tableau have developed numerous integrations, many of which include Salesforce's Einstein Analytics and its AI capabilities.
The vendor made Pulse, a tool that uses generative AI to automatically surface and summarize insights in natural language, generally available on February 22. Einstein Copilot in Tableau, meanwhile, is the evolution of Tableau GPT and represents an integration between Tableau and Salesforce's Einstein Copilot, which itself was built on an integration between Salesforce and OpenAI.
Tableau is not the first analytics vendor to introduce generative AI copilot capabilities. For example, Microsoft has been integrating copilot capabilities throughout its offerings and MicroStrategy made its AI assistant generally available in October 2023 before unveiling an embeddable version on March 26.
However, like Einstein Copilot for Tableau, most of the AI assistants that have been introduced remain in some stage of preview or testing, noted Doug Henschen, an analyst at Constellation Research. Even those that are generally available are so new that it's difficult to gauge how one performs relative to another.
Meanwhile, the security of language models and the cost of building and running generative AI applications are concerns for analytics consumers, leading many to wait before adopting new tools, Henschen continued.
"Tableau is in the middle of the pack in terms of the timing of its GenAI capabilities, but that's just fine with many customers," Henschen said. "I'm still hearing a lot of GenAI skepticism and wait-and-see attitude when it comes to adoption."
Southard Jones, Tableau's chief product officer, acknowledged that Tableau has been slower than some other vendors to introduce an AI assistant. The vendor's pace, however, has been purposeful, he continued.
"Copilots have been out and I'd say they've been met with varying levels of acceptance," he said. "They haven't all lived up to the hype. We've spent quite a bit of time making sure that [Tableau's AI assistant] delivers value and meets the hype."
In addition to launching Pulse and unveiling the beta availability of Einstein Copilot in Tableau, Tableau in February released its initial 2024 platform update, including Pulse enhancements and Tableau Cloud on AWS Marketplace. The vendor's final 2023 platform update featured new embedded analytics capabilities.
Einstein Copilot for Tableau, a new AI assistant now in beta testing, enables Tableau customers to query and analyze data using natural language.
New capabilities
Einstein Copilot in Tableau is designed to assist data experts and existing business users, according to Jones.
While Pulse targets employees within organizations who might not have much experience with analytics by delivering and summarizing insights in natural language, Tableau's AI assistant was built to help those already using Tableau improve productivity and delve deeper into data than was previously possible.
Specifically, the AI assistant includes the following:
Recommended Questions, a feature that automatically analyzes data and suggests questions so users don't miss potential insights.
Conversational Data Exploration, a capability that understands the context of a conversational thread so users can not only ask questions of their data but also follow up with additional queries that dig deeper into data and lead to sharper insights.
Guided Calculation Creation, a feature that guides users through the complex process of writing the syntax needed to calculate KPIs and metrics.
Together, the features should help enable enterprises to broaden their use of data to inform decisions, according to Mike Leone, an analyst at TechTarget's Enterprise Strategy Group.
For the past two decades, only about a quarter of employees within most organizations have had the expertise to use analytics tools, primarily because of the need to write code to interact with data. Generative AI assistants such as Einstein Copilot for Tableau eliminate many of the complexities related to data exploration and analysis, opening use of BI platforms to new users.
"It's really about enabling the broader business to gain confidence in exploring and analyzing data," Leone said. "For a while, we've heard about democratizing data and analytics through self-service, but … it just hasn't come to fruition. [Now] an overwhelming majority of organizations are quite bullish on generative AI solutions finally enabling that desired level of democratization."
In particular, Recommended Questions has the potential to help users get more out of data exploration, he continued.
"I love the fact that it can quickly look at a dataset and provide recommended questions based on that data set, especially for folks that lack training or expertise," Leone said.
Henschen, meanwhile, noted that given that Einstein Copilot in Tableau not only provides conversational analytics capabilities but also recommends questions, it goes beyond the AI assistants some Tableau competitors have introduced.
"I've seen lots of GenAI features focused purely on natural language-driven query and data analysis. But Einstein Copilot in Tableau … promises in-product help, proactive guidance and the curation and automation of repetitive tasks," he said.
In addition, Henschen noted that Recommended Questions and Guided Calculation Creation are derived from Tableau's 2021 acquisition of Narrative Science, which was a prominent vendor of natural language processing and data storytelling capabilities.
That lineage leads to strong potential.
"I have high expectations, but we have yet to see a generally available product," Henschen said.
Leone likewise said that Einstein Copilot for Tableau seems more full featured than some of the AI assistants introduced by other vendors.
Most of them deliver similar value, he noted. And if vendors don't provide one, they risk losing out on potential new customers to vendors that do offer AI assistants. Meanwhile, there are capabilities that can help one stand out from the rest. In the case of Tableau's AI assistant, Guided Calculation Creation is one such feature.
"This one is quite interesting to me because it's an area that is a very challenging task for generalists," Leone said. "The fact that generalists are now empowered to use natural language to parse long string fields and extract subsets of information is a big deal."
Tableau is in the middle of the pack in terms of the timing of its GenAI capabilities, but that's just fine with many customers. I'm still hearing a lot of GenAI skepticism and wait-and-see attitude when it comes to adoption.
Doug HenschenAnalyst, Constellation Research
Beyond the three main features, Einstein Copilot for Tableau provides transparency, enabling users to see how responses were derived so they can check for accuracy and know whether to trust a given output, Jones noted.
In addition, the tool was developed using Salesforce's Einstein Trust Layer, which provides security measures when integrating with OpenAI. The Trust Layer forwards only metadata to OpenAI, masks any potentially sensitive data such as personally identifiable information, checks the relevancy of results, alerts users when results are toxic or biased, and suggests ways to rephrase queries that might lead to better outcomes.
Meanwhile, the beta testing process will be used to test both quantitative and qualitative measures, according to Jones.
Qualitative measures include working with Tableau customers to see whether they are getting value from the AI assistant and finding applications for its use. Quantitative measures include tracking whether beta customers are increasing their use of Einstein Copilot for Tableau, meeting performance benchmarks and ensuring data security.
"[Security] is showstopper for us. It's probably the number one measure," Jones said. "The biggest reason for going to beta is to make sure [Einstein Copilot for Tableau] is really, truly trustworthy."
Future plans
Since its release, Pulse has been the most quickly adopted tool in Tableau's history, according to Jones.
With Einstein Copilot for Tableau now in beta testing with general availability planned within the next six months, the vendor is making its next product development plans. More AI figures prominently, Jones said.
One feature on the roadmap is to integrate Einstein Copilot for Tableau with Tableau Prep. In addition, integrations between Pulse and other existing Tableau tools are planned. Customers, meanwhile, have requested ways for AI to help them configure metrics unique to their business.
"We'll be incorporating AI into more and more areas of the product that will help drive the efficiency and trust of our core users as well as reach more of those casual business users," Jones said.
Improving its AI capabilities and adding new ones is a logical area of focus for Tableau, according to Henschen.
Tableau Conference 2024, the vendor's annual user conference, starts at the end of April in San Diego. It wouldn't be surprising if Tableau unveiled its next wave of generative AI capabilities at that time, he continued.
"I'm eager to see the next steps for Tableau Pulse, Tableau AI and Einstein Copilot in Tableau," Henschen said. "I'm sure we'll hear more on all three fronts."
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.
