Tableau launches Pulse, a GenAI-fueled insight generator

A broader reach

Generative AI has been the dominant trend in analytics since OpenAI's November 2022 launch of ChatGPT, which was a significant improvement in large language model (LLM) technology.

Since then, generative AI has become the focus of product development at Tableau, as well as tech giants AWS, Google and Microsoft. Specialized data management and analytics vendors Alteryx, Domo, Informatica, Qlik, ThoughtSpot and many others have also launched GenAI initiatives.

By adding LLM technology to their platforms, data management and analytics vendors can finally enable true natural language processing (NLP), which potentially eliminates one of the major barriers holding back the widespread use of their platforms.

Data management and analytics platforms are complex, requiring code to carry out most tasks and data literacy training to decipher outputs. As a result, analytics use within the enterprise has stagnated at around a quarter of potential employees for about two decades.

NLP capabilities developed by vendors were an attempt to make analytics use more widespread, as were low-code/no-code tools. The NLP capabilities, however, had limited vocabularies that still required users to be data literate while the low-code/no-code tools were incapable of deep analysis.

LLMs, however, are trained on public data and have the same vocabularies as dictionaries. In addition, they are capable of inferring intent. As a result, they enable true natural language interactions and reduce data literacy requirements.

In addition, because generative AI tools can both generate code on their own and also translate natural language to code, LLMs can make data workers more efficient by lessening the time it takes to carry out the data preparation tasks that dominate their time.

Pulse aims to broaden the use of Tableau within organizations, according to Southard Jones, the vendor's chief product officer.

Despite Tableau's best efforts, the vendor's platform fell in line with those of other analytics vendors whose tools were used by some workers within organizations but not all, he noted. To make Tableau's use more widespread, the vendor sought ways to engage potential users in different ways.

Jones noted that Tableau succeeded in engaging people who were data curious, such as business analysts. But the vendor did not succeed in engaging everyone in an organization that relies on statistics, such as frontline workers whose sales figures could affect salaries and bonuses.

To reach others, Tableau conducted research and determined it needed to come up with a design that reduces reliance on dashboards, enables users to engage with data using natural language and reaches users where they work, rather than in a BI environment.

Toward that end, Tableau acquired data storytelling and NLP specialist Narrative Science in December 2021. In addition, it developed the idea for Pulse.

"Pulse will enable people who may not be using Tableau a lot today to use data in their day-to-day life," Jones said.

Given that potential outcome, Pulse is an important addition for Tableau beyond simply being its first generally available generative AI tool, according to Leone.

"Organizations are chomping at the bit to leverage GenAI capabilities with analytics environments," he said. "This will enable organizations to make significant progress in enabling the wider business to experiment with and analyze data on their terms."

Pulse in action

Using generative AI, Pulse can monitor Tableau customers' data beyond what even a team of humans are capable. When it detects metric changes and new trends that humans might not otherwise discover, the tool is then able to alert users in natural language and provide accompanying data visualizations.

In addition, Pulse explains the drivers behind changes and trends so customers can more fully understand what is happening in their business.

Those insights, embedded into users' workflows rather than delivered solely in Tableau's environment, then enable customers to ask follow-up questions and take data-informed actions.

That emphasis on insight generation and monitoring metrics, meanwhile, differentiates Pulse from the generative AI tools introduced by Tableau's peers, according to Menninger.

While tech giants including AWS, Google and Microsoft have all unveiled chatbots and AI assistants that enable natural language query, Pulse is instead focused on metric analysis with generative AI used to provide narrative context.

"All these things are helpful to users, and it will be exciting to see other ways in which GenAI is applied as well," Menninger said.

As part of Tableau's effort to provide insights from Pulse within users' normal workflows, the tool includes Pulse on Mobile. The feature alerts users about metric changes and trends on their mobile devices, including within Tableau's mobile app and email.

In addition, Pulse includes Pulse Slack Digest, which delivers generative AI-generated summaries and contextual insights related to metrics and trends within users' preferred Slack channels.

Beyond delivering insights to Tableau customers in their preferred workflows, Pulse includes a Metrics Layer that enables users to create standard metric definitions by using metadata to determine the meaning of terms and the business context of metrics.

By adding Metrics Layer, Pulse aims to ensure that there is only one metric for a set of metadata so that there is consistency across an organization. The feature prevents different users or departments from giving different definitions and descriptions to what might be the same metrics.

"There's a lot to like between the insights platform, mobile and slack integration, but I think the Metrics Layer will be incredibly valuable," Leone said.

He noted that without understanding the business context of potential insights, users are unable to take actions based on those insights.

"The Metrics Layer will help customers address that challenge head on," Leone said.

While Pulse's development stemmed largely from Tableau's effort to expand its reach beyond a small segment of potential users, Jones noted that customer feedback also played a role.

Just as Tableau wanted to engage more people within organizations, many customers wanted the same thing. They didn't ask for Pulse specifically. Instead, they asked for things such as dashboards that are easier to decipher.

"They asked us for many different things that were code words for, 'How do we engage more people?'" Jones said. "So, they've asked for this, but not this specific feature. What they really asked for was something more."

He added that Tableau put significant time into research and development to get Pulse ready for general availability, including a three-month beta testing period with about 1,000 users to help refine the tool.

Next steps

Unlike Tableau's overall platform that is updated three times per year, the vendor plans to update Pulse every two weeks, according to Jones.

"This is a capability that we have to continually iterate to see how we're engaging users and get them to digest information and make decisions on it," he said.

In its initial form, Pulse will deliver a headline alerting users to insights, a description of metric changes and trends and information about what is driving changes and trends. In addition, the tool enables users to ask basic follow-up questions in natural language.

Tableau's near-term focus -- the first few biweekly updates -- will be on enabling deeper follow-up analysis with more extensive question-and-answer capabilities, according to Jones.

Accurate question-and-answer capabilities are difficult to get right, given generative AI's propensity for delivering sometimes wildly incorrect responses called hallucinations, he noted.

"If you look at the products that are out there that enable truly ad hoc Q&A of data, you get a lot of hallucinations," Jones said. "We're working hard to release an ad hoc Q&A version that will be accurate and trusted."

Menninger, meanwhile, said that because of Pulse's potential to make analytics use more widespread, there remains an opportunity for Tableau to focus on improving productivity with future generative AI development.

In particular, more efficient data preparation could be a target for Tableau.

"Since so much of the time spent on analytics is in the process of data preparation, I expect we will see GenAI used to dramatically simplify the processes," Menninger said. "But there are many areas where GenAI can improve productivity, ranging from SQL generation to self-documenting analyses. I'd encourage Tableau and other vendors to pursue all these opportunities."

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.