Get your FREE 30-day trial.
Please complete all fields.
Did you know more than half of organizations report using 800 or more different applications? With employees working across so many tools, there is no central place to surface insights or embed intelligence. You’re not going to put those insights to use if you have to open a new window to get to them — that’s why we’re so excited to bring Einstein Discovery to you, wherever you’re working.
The Einstein team here at Salesforce believes in empowering everyone — from CEOs to customer service agents and sales managers — with AI-powered insights, predictions, and recommendations. Today we announced new AI services to further that mission by empowering admins and developers to build custom artificial intelligence with just a few clicks or simple code. Most exciting to me though, is the Einstein Predictions Service. For the first time, developers and Salesforce admins can embed AI-powered insights from Einstein Discovery into any third-party non-Salesforce application such as human resources (HR) or enterprise resource planning (ERP), bringing predictions to the point of decision making.
Insights accessible anywhere
Einstein Discovery analyzes millions of rows of data within minutes to discover patterns a human would never be able to find, and surfaces AI-driven predictions and recommendations. And now, even if you don’t use Salesforce regularly, you can still benefit from Einstein’s AI-powered insights.
Here’s how: an admin or developer builds a prediction within Einstein Discovery in a few clicks, utilizing Salesforce data and external data sources. After assessing the model metrics, they can now embed an Einstein component in a third-party app to display the insight in a location other than Salesforce. For example, Einstein Predictions Service lets an HR manager see the likelihood that an employee may leave their job directly within an HR system, like Workday. This means that the HR manager does not need to navigate to a new system or open a new window — Einstein’s prediction is embedded directly where they’re working, with the insight they need, with no disruption to their workflow.
Another instance could be for supply chain managers at a medical device company. An admin or developer would build a model within Einstein that combines account information from Salesforce with order information from an ERP system. Einstein Discovery is now able to surface predictions on increasing demand for devices and make future manufacturing decisions within the ERP system where supply chain managers are already working.
Einstein Discovery is an intelligent data discovery feature to surface AI-powered insights. As we are enhancing the ability to build custom predictions and extend them outside of Salesforce, it’s never been more important to make sure it’s done in a trusted way. Salesforce is a values-based company, with trust and equality being two of our core pillars, so we’re empowering users to build AI with these values in mind. Not only is Einstein Discovery a one-stop-shop for building your custom AI, it also provides alerts and guardrails to help ensure you’re building trusted AI that is transparent, responsible and accountable:
Transparent
Einstein Discovery shows you the top predictive factors for every prediction. For example, if you build a prediction for churn, your top predictive factors might be customer tenure, products purchased, number of service cases opened, or lifetime value. By sharing these factors, the users consuming the prediction can better understand and trust the recommendations and the actions driving them.
Additionally, for more advanced transparency, your company’s data science team can access the underlying R code of models to perform an independent validation based on your organization’s own ethical standards. This is like looking under the hood to see how everything is working for analysts that do understand data science and may want to make manual changes to the model as needed.
Responsible
As AI is being implemented at record speeds compared to past technologies, we need to be aware and cautious of the potential bias that may come from skewed data sets. Potential bias varies on a case by case basis, so customers can create protected fields such as gender or race that should not impact predictions. Einstein Discovery will then warn users if there is a possibility of bias in a dataset via a pop-up alert. For example, if race is a protected field for your model, you’ll be alerted that zip code may also be a proxy for race, and receive a recommendation to remove it from your model.
Accountable
Model metrics show how your models will perform before, during and after you roll them out to assess the quality of the model and understand the accuracy of the predictions. Then, a feedback loop allows business users to weave their expertise into the model to continually improve it. Users can provide direct and indirect feedback to models to constantly update and become more accurate. For example, if Einstein assigns a high score to a lead that doesn’t end up converting, the model will learn from that indirect feedback to tune itself and adjust for future predictions.
Thresholds can also be customized depending on the prediction. Low-risk recommendations, like what marketing email to send next, can have a low threshold for success. High-risk predictions, like a new sales strategy for a big prospect, require a much higher threshold for accuracy.
Be a data Trailblazer
With the latest AI services, we’re making it easier than ever for admins and developers to build and extend predictions to everyone in a trusted way. Einstein Predictions Service further democratizes AI by extending intelligent experiences, powered by Salesforce, to any apps that make sense for our customers. Earn one of the Einstein Analytics superbadges for Data Preparation or Insights, and become an Einstein Analytics and Discovery consultant with the first ever Einstein certification to take the next step on your own AI-augmented analytics journey!
