Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.
Salesforce’s Tableau arm is making a case for employing AI to drive a new era of business science using analytics applications infused with machine learning algorithms the average end user can easily employ.
Tableau has added support for Einstein Discovery to the 2021.1 update of its namesake analytics application. Originally developed by Salesforce, the Einstein Discovery module uses machine learning algorithms to surface patterns in data. Including that capability within Tableau will make it possible for business users to employ data science techniques to analyze data without having to write code and without any intervention on the part of a data scientist team, Tableau CTO Andrew Beers said.
Prior to being acquired by Salesforce in 2019, Tableau had begun to embed machine learning algorithms within its application, along with natural language processing (NLP) capabilities. The merger with Salesforce enables Tableau to leverage Salesforce’s ongoing Einstein research and development efforts.
As analytics applications continue to incorporate capabilities such as Einstein Discovery, organizations will need to determine when the AI model that needs to be constructed is simple enough for end users to create using a graphical tool versus adding yet another project for a data scientist team to complete. “That line today is kind of blurry,” Beers said.
Data scientists are, of course, in short supply. And most data science teams have a backlog of projects they are not likely to complete anytime soon. Most are fortunate to be able to successfully complete more than a few projects in a year. However, the backlog of projects a data science team is being asked to complete might decline as more AI capabilities are added to analytics applications.
The bulk of AI models end users might ask a data scientist to build would never be built in the first place, as there simply isn’t enough time. The only way to address those requirements is to democratize AI within the context of an application such as Tableau. Those efforts will lead to a set of best practices for business science that will be distinct from the more complex AI projects a typical data science team will take on, said Beers.
It’s not quite clear what level of expertise will be required to enable end users to build their own AI models. Einstein Discovery comes with several built-in capabilities for surfacing bias and promoting AI model transparency. But most organizations would be well-advised to review AI models created by end users before making any business decisions that cannot be reversed.
In the meantime, organizations should expect AI models to become a lot easier to create using natural language processing (NLP) engines that are increasingly providing support for various speech recognition engines. Business users will be able to invoke those engines to create AI models that would, for example, make it much simpler to engage in what-if analysis involving multiple scenarios. That kind of capability is more valued than ever in the wake of a pandemic that proved conclusively that all business assumptions are subject to rapid change.
In fact, as the pace of change continues to accelerate in the age of digital business, it’s doubtful the average user will be able to keep pace without some AI help. That doesn’t mean AI models will replace the need for business analysts, but it does mean the days when data science required a small army of specialists are coming to an end as the AI playing field becomes increasingly level.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.