Tableau has made substantial investments in AI, acquiring a number of natural language and automated analytics companies. Given that Tableau was acquired by Salesforce, users now have access to Salesforce's powerful Einstein AI engine.
Experts said there are several ways Tableau AI features can complement Salesforce today. There are also several gaps in both tools that could be addressed by new features down the road.
"Salesforce has been clear that they want to accelerate Tableau's mission to help people see and understand data," said Tal Doron, associate vice president of solution architecture at GigaSpaces Technologies, an analytics processing platform. "They have so many AI technology assets; it is very likely that they will use these to evolve Tableau offerings faster."
The entire market is trending toward more advanced analytics, where AI and machine learning are blended with traditional analytics and BI capabilities.
"You'll see this in predictions being surfaced in analytics but also in terms of how we explore data and ask questions of our analytics solutions," said David Dixon, director of analytics at Atrium, an AI consultancy.
Great examples of this are the Explain Data and Ask Data features already available in Tableau and Einstein analytics. With these capabilities, users can have their questions answered simply by asking the questions via natural language processing (NLP) interfaces.
"It's like Siri for analytics and BI," Dixon said.
Dixon sees gaps in both offerings that are filled by the other.
Tableau leads with on premises, behind the firewall data and non-CRM/Service data sources and use cases. It's good at connecting and extracting data from a variety of on-premises data sources and can combine data from multiple systems. This makes it a clear choice for operational and back-office analytics.
Tal DoronAssociate vice president of solution architecture, GigaSpaces Technologies
"However, if you need actionable analytics and insights, especially predictive insights, surfaced for your CRM or Service Cloud users, Einstein is still the better choice," Dixon said. Salesforce excels at analyzing customer data to determine the next best action for sales agents.
Another strength of Tableau is the ability to analyze data more intuitively to ask the bigger questions, such as, "What's contributing to customer churn?" These kinds of Tableau AI features can help to bring explainability to AI and decision learning models.
"Capabilities such as [Tableau's] Explain Data can evaluate every dimension in the data set automatically and provide visual results that can be queried with a simple click, resulting in more meaningful insights faster," Doron said.
For example, Tableau's Explain Data can be used to discover that a high volume of mobile phone sales contributed to an unusually high level of profitability in New York state. Explain Data delivers explanations in the form of interactive visualizations, which users can further explore by clicking on different parameters.
Doron said other Tableau AI feature strengths include automatic clustering and NLP. Automatic clustering uses a k-means clustering algorithm to identify significant groupings of data. Tableau uses NLP to analyze and predict a search or filter and natural language generation to automatically create verbal descriptions of visualizations.
Other experts are concerned that businesses could run into problems when attempting to visualize complex AI. Business users who are accustomed to cut-and-dried answers derived from straightforward statistics risk drawing erroneous conclusions through poorly thought out AI analytics.
"Unlike a simple formula for revenue minus expenses, the results of AI models need experienced user interpretation to be meaningful and fully understood, " explained Malcolm Thorne, venture capital partner at 4490 Ventures.
These AI results can also change daily as the data changes. The potential validity and reliability can change frequently based on what data is being processed. A simple formula for calculating net operating income does not change ever, unlike more complex AI models.
"Hence it is unlikely that firms will be able to enable mass visualization of more complex AI techniques easily," Thorne cautioned.
There are also some significant gaps in both tools at the moment.
"Neither tool is very good at clustering algorithms nor doing things like helping you group similar records together," Dixon said.
For example, if a user wants to answer questions such as, "What other products did clients similar to this client purchase?" neither Einstein nor Tableau does that particularly well.
However, both products can be extended to answer those questions via other AI platforms, such as Amazon SageMaker, or with custom models in Heroku.
"Modeling capabilities that enable product recommendations or next best product would be a great addition to both tools," Dixon said.
What could be coming
Andrew Beers, CTO at Tableau Software, was cautious in laying out the roadmap for Salesforce and Tableau AI features other than to say, "There are more opportunities to integrate than we can list. From a business and a technology perspective, we think we will truly change the game in terms of how to make data-driven decisions easy for everyone in an organization."
Meanwhile, Salesforce has announced it is further developing AI capabilities for voice analysis.
"Analyzing voice communications could add another dimension to analytics related to customer sentiment and retention," Doron said.
Also, the companies are investing in a variety of natural language capabilities that could enhance various aspects of Tableau AI features.