Thinking back to Dreamforce ‘19, artificial intelligence (AI) took center stage. From new feature launches, to discussions about how to create more ethical models, Salesforce leaders made it clear that AI is at the heart of its innovation plans for the next decade.
Here’s what you need to know about all the talk about AI from insights we picked up at Dreamforce and wider – and how Admins can expect to see Salesforce AI tools making waves in 2020 and beyond.
Einstein Sharpens Its Voice
One of the most exciting announcements during the Dreamforce keynote speech was that voice technology is poised to have a huge impact on the way users interact with Salesforce.
Einstein, Salesforce’s AI platform, is launching several new voice capabilities.
For customers, the debut of Call Coaching promises insights that can boost close rates and deliver the information that matters most to them. A similar feature called Service Cloud Voice can also transcribe calls in real-time to offer recommendations for sales reps to help serve their customers better and faster.
For Admins, this means less data entry and fielding fewer technical and numerical questions from users so that they can spend more time doing the meat of their job: helping users get the most out of Salesforce.
Users can now ask Einstein direct, plain-speech questions, like, “how many deals were closed this quarter?” The query gets picked up through a smart speaker that can be placed on a desk. The process is delightfully simple – and means a user doesn’t have to track down an Admin to get the data they want.
Anything that requires domain expertise, though, won’t be handled well by voice. Einstein is better for quick numerical or technical questions. For users who are having trouble interpreting data, Admins should consider tools that translate data points to stories told in plain English.
The key value an Admin provides will always be data context and best management practices; what artificial intelligence offers are better ways for users to navigate Salesforce. With Einstein and other AI tools onboard, Admins can focus on making Salesforce a more powerful tool for their users.
AI Drives Smarter Predictions
There was a lot of buzz around predictive analytics at Dreamforce. Salesforce continues to blaze ahead of its competitors: it’s the first CRM platform to introduce AI-enabled predictive analytics, and its integrations will start rolling out in 2020.
One surprise announcement from the event may be of particular interest to Admins: beginning next February, Salesforce will offer one free prediction for every organization with an enterprise license to Einstein Prediction Builder. This Einstein tool gives custom AI to Admins, allowing them to build prediction models without writing any code.
For the vast majority of Salesforce users, though, predictive models are brand new. Business executives are familiar with descriptive analytics, or what data tells us about what’s happening now. Venturing into the world of predictive analytics, there will no doubt be lots of questions about a prediction’s confidence.
Admins will have to understand how Salesforce models make these predictions so that they can explain to users why – and whether – we can trust them. They’ll need to understand the scope and limitations of their datasets, too, in order to choose and implement the models that will be the most insightful for their organizations.
Toward More Ethical AI
How can we build algorithms without bias? We expect machines to be logical, but they learn what we teach them, which often includes our own biased ways of thinking.
There was a large – and deserved – focus around “ethical AI” at Dreamforce. Salesforce’s Chief Ethical and Humane Use Officer, Paula Goldman, gave an excellent presentation on the ethics of the use of AI and shared her insights on how Salesforce developers can make machine learning models that reduce bias.
What are the stakes here? Take, for instance, this Reuters story from last year: Amazon’s HR department was using an AI resume review tool that categorically discarded female applicants. The machine had been taught to penalize resumes that included words like “women’s,” such as “women’s soccer captain.” A model designed to be an objective and efficient hiring tool instead reinforced discriminatory recruiting patterns.
But if you’re using anti-bias strategies such as peer review, inclusive model design, and more, a system bias that might go unnoticed by one developer can be flagged by another.
Building ethical AI models isn’t just the task of developers, though. Admins should strive to be more mindful about the ways they interact with and influence these models, too.
As more AI models make their way into the Salesforce toolkit, Admins need to understand why these models make certain decisions and exercise constant skepticism about the data they work with. Admins should strive to see no person and no piece of data misjudged or misrepresented.
Summary: Salesforce AI Is Just Getting Started
If there was one major takeaway from Dreamforce, it’s that Salesforce is all in on AI.
On the organization level, though, the great potential for AI won’t be realized all at once. Organizations will likely test one AI tool at a time. You need to prove that you can get value out of a single AI initiative, and then you can justify its use beyond that initiative.
As ever, Admins need to be fluid, adaptable, and continuously learning about Salesforce’s changing AI capabilities. After all, they’re the ones who bring the trailblazing happening at Salesforce HQ out into the world.