Here’s the list of Einstein features that help make Salesforce the world’s smartest CRM.
Want to find stuff faster with search results that are tailored to the unique way you work in Salesforce? Search personalization is here. It helps you cut through the clutter with search results based on what’s most important to you, including geographical locations, industries, statuses, product areas, and people.
Enhanced instant results turns the global search box into a supercharged productivity hub. Click in the search box to instantly access record previews, page-level record actions, related list quick links, and suggested searches.
Can’t find a report or don’t have time to create one? Conversational search gives you instant access to important data by turning your search terms into record filters. Just enter search terms the way you start a conversation. For example, enter “my closed cases this month” to see a list of your recent cases with a closed status.
Recommended Result gets you to the right record faster when we’re confident we know what you’re looking for. It’s like putting purple stripes on your luggage at the airport carousel to make sure it stands out from the crowd.
Search results should show what matters most. If you’re in sales, knowing the account owner and the industry is critical. If you’re a service rep, the account’s support level is key information. With profile-specific layouts, you can fine-tune search results layouts for an object for each unique profile in your org. Profile-specific layouts are supported only by objects with customizable layouts. Users who don’t have a profile-specific layout assigned to them see the default search results layout.
If you use a quarterly forecasting schedule, you can now use the power of Einstein to improve forecasting accuracy, predict results, and track how sales teams are doing.
Does your sales team need lead scores for only some leads? Tell Einstein to score only leads that contain certain values in lead fields. For example, if your sales team wants to score only leads from the communications industry, tell Einstein to include only those leads.
Opportunity Contact Role suggestions now respect the field-level security from the Contact object. Sales reps without access to contact fields, such as Email, Title, or Phone, no longer see those fields in the Einstein component or list views. Also reps don’t see contact field values for contact records they don’t have access to. Lastly, when viewing the detail page for Opportunity Contact Role suggestions, contact fields don’t appear. To see the contact fields, navigate to the contact record.
Get more control over whose email data is captured. Use the Excluded Addresses list to prevent data from syncing. Plus, let reps sync event series.
Use standard reports to analyze opportunities based on opportunity scores. Previously, you could use opportunity scores with only custom report types.
Get a clearer picture of what drives your opportunity scores. The factors used to create the opportunity scoring model are now available in custom report types.
We introduced two new objects for Einstein Opportunity Scoring. Use the SalesAIScoreCycle and SalesAIScoreModelFactors objects to retrieve information about opportunity scores and their factors.
Einstein Article Recommendations uses data from past cases to identify Knowledge articles that are most likely to help your agents address customer inquiries.
Build bots that clean up after themselves! Use an End Chat rule action to add conditions that help the bot sense the natural end of a conversation and automatically close the session on behalf of the visitor.
Bring your own natural language processor (NLP) into Einstein Bots to create a multilanguage experience or to tie into your existing systems. Intents and utterance data are passed through Apex to the processor of your choice. A two-way system connects your NLP to Einstein via an Apex template, so customer inputs are sent from your Einstein bot to your provider. The bot can also receive intent and entity information from your third-party provider to use conversation routing.
Bots and agents work better together, and we’ve made transfers even smarter by checking the availability of agents before a transfer takes place. Define custom messaging to keep the customer informed as to the status of the transfer.
Get the most helpful Einstein Bot content at your fingertips with a carousel on the Setup screen. Get access to bot recipes, troubleshooting tips, and best practices.
You can write prediction scores automatically to selected Salesforce fields without coding. Easily integrate predictions without involving Process Builder or a managed package with a trigger.
Previously, Einstein Discovery relied on regression models to predict outcomes. Einstein Discovery now adds a second type of model that is based on a prediction optimization approach known as gradient boosting learning algorithms. When you create a story, Einstein Discovery generates predictions using both types of models and shows the results of the model that performed better. You get the best of both approaches.
Determine the accuracy of your logistic models by visually comparing predicted outcomes with actual ones. Then use this feedback to fine-tune your model and produce better predictions. An actual outcome is data that is not expected to change because it has reached its terminal state. An example of finalized data is the date on which an order shipped. Define the conditions under which your story’s outcome variable has attained its terminal state. That way, Einstein Discovery knows which outcomes to include in the performance analysis.
In Model Metrics, you can now display the metrics for multiple models side by side. See how model metrics stack up against each other. Compare segments to reveal the most important variables in each segment. Use what you learn to improve your models and achieve better predictions.
When developing a model for a categorical field, you can set an optimal threshold that represents the cutoff for the two buckets you are predicting. For example, you can specify a cost ratio between the false positives and false negatives. Then let Einstein Discovery pinpoint an optimized threshold for the business case associated with your story. The threshold value represents the tradeoff between the true positive and false positive rates.
Let Einstein Discovery select the best data to analyze for your story’s goal. It searches your dataset, chooses the columns that correlate to the outcome, and excludes the columns that have no correlation. After your story is created, you can manually change the column selections.
For logistic regression models in which the outcome variable is a text field, a new residuals plot chart reveals the robustness of your model. A residual represents the difference between the model’s predicted value and the actual outcome value. An actual outcome is data that is not expected to change because it has reached its terminal state. An example of finalized data is the number of items a customer received in a shipment. Define the conditions under which your story’s outcome variable has attained its terminal state. That way, Einstein Discovery knows which outcomes to include in the plot chart.
Models built with biased data can produce biased predictions. Disparate impact is one example in which data reflects discriminatory practices toward a particular demographic, such as gender disparities in starting salaries. Einstein Discovery alerts you to variables that are being treated unequally in your model. You can remove disparate impact bias from your predictions to produce more ethical and accountable models.
If all you want from your data are What Happened insights, you can skip predictive analysis of your dataset. Story creation is faster because Einstein Discovery doesn’t generate predictions and improvements. After your story is created, you can manually add predictive analysis if you change your mind.
When data changes in the source dataset, you can now choose to analyze the updated data instead of the snapshot taken when the story was created. When you open a story, Einstein Discovery notifies you when changes have occurred to columns or rows. Previously, a story was always pinned to the original snapshot of the data.
Einstein Data Insights limits have increased. You can now create insights from reports containing up to 500,000 rows and 50 columns of data. In addition, you can create up to 1,000 Einstein Data Insights analyses per org per day.
Einstein Discovery can now analyze Einstein Analytics datasets with row-level security predicates and sharing rules that are associated with Salesforce sharing inheritance. All users who access the story can see the results of the story. They don’t need the same row-level access as the story creator. Previously, you needed the “Ignore predicate when creating story from dataset” permission, which is deprecated in the Winter ‘20 release.
Customize the look of Einstein Discovery Predictions embedded on Lightning record pages. For logistic regression models (binary classification problems), specify labels that appear when the prediction is higher or lower than the model threshold. Examples: win or loss, retain or churn, and so on. Filter recommendations on how to improve a prediction. Set the maximum number of recommendations or show recommendations that impact the outcome by a minimum percentage.
We plan to retire Einstein Discovery Classic in Spring ‘20. Current Einstein Discovery Classic users need the Einstein Analytics Plus license (required for Einstein Discovery in Analytics) to recreate datasets and stories in Analytics Studio. Einstein Discovery Classic will be replaced with the new experience in all Developer Orgs with the Winter ’20 release.
When you build a prediction, you no longer need a field that answers your prediction question. As long as the records on the object that you base your prediction on have the data, you can use filters instead.
Instead of filtering only on absolute field values, you can now filter on the value of one field compared to the value of another field or on a point in time. Make your filter logic more meaningful and relevant to your prediction question.
Help your service agents take action quickly. Call an autolaunched flow to update records or send an email behind the scenes via a recommendation.
Create strategy templates that your subscribers can customize and build on. Share them in managed packages that you publish on AppExchange. A managed package can contain both strategy templates and strategies protected as your intellectual property (IP). Subscribers can open a template in Strategy Builder and clone it to customize for their own use. Strategies not marked as templates are IP protected and can't be edited or cloned. You can upgrade strategy templates and IP-protected strategies as part of a package upgrade. When you push upgrades to strategy templates, you don’t affect subscribers' copies.
Accidents can happen when you create strategies. Now you can easily undo and redo changes, like mistakenly deleting a parent element or moving an element from one branch to another.
Create expressions more quickly and accurately when using a Branch Selector element to branch recommendations. No more manually entering picklist values. Now you can select the values, and Next Best Action populates the expression for you.
We added titles to the list of attributes that you can show for a recommendation in the Next Best Action Lightning component in Lightning App Builder or Communities. Add a title to a recommendation so that your agent or user can easily identify it.
Add the Next Best Action Lightning component to an app’s Home page to display an aggregated set of recommendations. For example, show an agent a list of key accounts to follow up with after a specific number of days has passed since the previous contact.
The response for Einstein Vision API calls that return model information now contains the language and algorithm fields.
The response for Einstein Language API calls that return model information now contains the language and algorithm fields.
Get details on the logic behind Einstein predictions via insight objects. When an Einstein feature, such as Einstein Prediction Builder, makes a prediction and saves the results, an AIRecordInsight record and several associated child records are created. Use these records to understand how Einstein predictions are made and apply custom logic after the predictions are saved to improve and customize predictions.