Salesforce Einstein: Bots Faster to Set Up, Data Checker in Prediction Builder, and Sales Cloud Quarterly Forecasts
When an opportunity score isn’t shown to sales reps because they have limited access to Einstein Opportunity Scoring, the score value is now marked as hidden.
Einstein intelligence is built on data. And the more data Einstein has, the more powerful the predictions are. But not everyone has enough data to build a predictive model of their own. To give more customers meaningful artificial intelligence, we’re building global predictive models that all customers can use. Global model looks for aggregate, anonymous trends across many Salesforce customers, so the first step is to gather data from multiple Salesforce orgs. We’ll use the data to build global models.
It’s now easier to create and edit Einstein Activity Capture configurations. Decide what it means to share activities with everyone. And, Activity Metrics is generally available with some key improvements.
The Smart Email Matching option that helps improve opportunity contact role suggestions is no longer available.
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.
Improve your forecasting model by sorting your opportunities into relevant groups, such as by region or product line.
Not only is Pardot Einstein Campaign Insights now generally available, we made it better. In addition to prospect details and list email engagement, find insights for other marketing assets. Get a bigger picture view with new explanations that feature activity around marketing forms and landing pages.
When a prospect shows all the right buying signals, you don’t have time to waste with administrative tasks, like assigning leads and posting to Chatter. Use Einstein Behavior Scoring as part of your rule criteria in Process Builder and Workflows to automate the day-to-day. With these tools, you get the best leads to your sales team faster.
Transfer each bot session to the right agent or queue by overriding the default queue on Chat buttons or the channel ID for Messaging. Create rule actions that route conversations to agents in specialized queues.
Spot each conversation’s turns, references, and rule conditions with an easy-to-use visual representation. Build and create dialogs in Dialog Details, then switch to Map to see the conversation’s configuration and framework. Dialog Map is read-only, so you can’t create or update dialogs from it.
Jumpstart bot implementations without adding 20 utterances or turning on Einstein to create an intent model for your bot. On dialog intents, you can choose to enter a few keywords that exactly match utterances used by customers. This “exact match” option helps you test and deploy simple bots faster. Exact match ignores capitalization but doesn’t overlook spaces or punctuation.
Expand your bots’ capabilities and give them access to features like Field Service Lightning by assigning them custom profiles. Each bot now has a Bot User, which specifies what a bot can do based on licenses, permissions, and objects. By default, each bot has a Basic Chatbot User profile, but we recommend that you create a custom profile to broaden your bot’s behavior.
Combine recommendations from Einstein Next Best Action with other guidance, and show it all in a single place: the Actions & Recommendations component. Create actions and offers as recommendations. Refine and personalize them with action strategies that use business rules, Einstein predictive models, and other data sources. Present the best options with other actions so that your workers can find the right steps fast.
To simplify the Case Classification setup, we streamlined some options and removed the Optimize button.
Combine the top recommendations from your Einstein Next Best Action strategies with other next steps that your agents can take. When your strategies produce recommendations, they appear in a tab on the Actions & Recommendations component. You no longer need a separate Next Best Action component to show them to your agents.
Create accountable AI models that detect and flag potentially biased variables. With protected fields, you can declare protected values to exclude from your models and receive notice of problematic correlations.
Use the new Overview and Prediction Examination tabs to get more feedback about your model so that you can refine it to get better predictions.
Einstein Discovery stories are now included in full copy sandboxes. Copy a story and the data to a sandbox to test it. No need to recreate the story from scratch.
Get help to optimize data analysis in your story with the Story Setup wizard. It shows you how each field correlates to the story outcome, identifies redundant or outlier data that you can exclude, and suggests bucketing improvements.
You can now write scores automatically to selected Salesforce fields. Easily integrate predictions without involving Process Builder or a managed package with a trigger.
Add Einstein Discovery predictions as a standard Lightning component in any record detail page. In the Lightning App Builder, drag the new Einstein Predictions component onto your record page. Then simply choose a prediction definition, set prediction units, and select display settings. On the record page, predictions are updated in real time, and no writeback to Salesforce is necessary.
Better bucketing improves your story insights. Einstein Discovery uses an unsupervised learning algorithm that analyzes numeric columns in your story. It suggests better buckets to use, such as different ranges for age or postal code data.
Calculations often result in fractions. But certain predictions make sense only when expressed as whole numbers, say the number of orders per month or the number of customers entering a store. Display them as whole numbers using automatic rounding.
Determine the accuracy of your model by comparing predicted outcomes with actual ones. Then use this feedback to fine-tune your model and get better predictions.
For Einstein Analytics dataflows, to score a dataset, add the new edScore transformation in the dataflow editor. Specify the name of the prediction definition to use and the name of the new dataset column to contain the score.
Einstein Discovery now generates stories with numeric outcomes faster by selecting just the features and variables required to build the model. The story creation process is streamlined and quicker. Optimized feature selection applies to regression stories but is not for classification.
Don’t have a lot of data? No problem. Descriptive insights require only a minimum of 50 rows of data. Predictive insights require a minimum of 400 rows of data. Select the dataset during story setup. If it contains fewer than 400 rows, Einstein Discovery generates insights for descriptive insights but doesn’t generate a model for predictions and improvements.
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 Summer ’19 release.
After deploying models with Einstein Discovery, use the Einstein Predictions Service Scoring API to embed your predictions into any website or application. Use the Scoring API REST endpoint to access the deployed Einstein Discovery predictive modes. Send model input variables, and get back predictions, reasons for the predictions, and recommendations on how to improve the predictions. Via embedding, you can score records located inside or outside of Salesforce.
Don’t wait until you’re done setting up a prediction to find out you don’t have enough records in your dataset or that the prediction quality is low. Einstein can determine how many example records and how many true and false values you need to make a useful prediction. Use Data Checker to get this information as you go. Find out early on that you don’t have enough records in your filtered dataset so that you can adjust your segment or example filters, and then recheck your data.
When you build a prediction, you use filters to define the set of records in your data set that your prediction is based on. Now you can apply custom logic to your filter conditions to be more precise about the set of records that you want to include.
Einstein Next Best action now lets you create expressions for filtering recommendations more quickly and accurately. Expressions use resources, operators, and values, and you can build them in two different modes: simple and advanced. Simple is declarative: just select or search to build your formula. Need a more complex expression? Advanced is the way to go: type your expression following the formatting and syntax guidelines you’ll find on the screen.
Now, instead of relying on static, pre-created recommendations, you can get dynamic recommendations and AI-driven predictions using two new elements: Generate and Enhance.
The new Map element lets you use formulas to create new fields and modify existing fields without having to use Apex code.
Packaging lets strategy builders distribute their Next Best Action strategies easily and at scale. Enterprise developers can package strategies for use in multiple Salesforce orgs, and independent software vendors (ISVs) can add strategies to AppExchange for distribution to their customers. You can package your strategies and certain dependencies and distribute them in both managed and unmanaged packages. Change sets are also supported.
Einstein Prediction Builder generates probabilities and predictions, like how much a customer is likely to pay for a service. When creating a custom formula field, you can now reference AI prediction fields.
Training requests from customers with a paid plan are prioritized before requests from customers on a free plan. A training request is any call to a /train or /retrain resource. You could experience a delay in training if you’re on the free tier of service and other requests are in the queue.
Each Einstein Platform Services account is now limited to 30 calls per calendar month to Einstein Vision and Einstein Language endpoints that return examples. This limit applies across all APIs that return examples. If you exceed this limit, you receive an error message.
Tap into the power of AI and train deep-learning models to recognize and classify images at scale. You can use pre-trained classifiers or train your own custom classifiers to solve unique use cases.
Use the Einstein Language APIs to build natural language processing into your apps and unlock powerful insights within text. The language APIs include the Einstein Intent API and the Einstein Sentiment API.