Key Takeaway: Amalgam believes that the go-live date of myEinstein will be the most important date for Enterprise AI in 2018 as it represents the day that AI will become practical and available to a broad business audience across industries, verticals, company sizes, and geographies.
On November 6, 2017, Salesforce [NYSE:CRM] announced the launch of myEinstein: services based on Salesforce’s Einstein machine learning platform to support point-and-click-based and codeless AI app development. This announcement was one of several new services that Salesforce built across platform (mySalesforce and myIoT), training (myTrailhead), and user interface development (myLightning).
myEinstein consists of two services: Einstein Prediction Builder to support the creation of custom AI models based on any Salesforce field or object and Einstein Bots, which can automate workflows associated with retrieving Salesforce information or answering Salesforce-based questions.
These services leverage APIs launched by Salesforce earlier this year: Einstein Language, which focuses on classifying customer intent and sentiment through language and through image identification; and Einstein Vision, which supports both image classification and object detection so that Salesforce developers can identify specific objects, object size, quantity of objects, and location based on the details provided in a picture.
Einstein Prediction Builder works as a click-based setup that guides users through the definition of the end goal that is being predicted, identifying core fields and variable drivers that the end goal is based on, training, and deployment of data scoring and predictive models. These automated data model building and customization capabilities were first announced in Salesforce’s Spring release but this announcement greatly democratizes access to this capability. The outcomes of these models can then be embedded into a Salesforce dashboard or report to provide users with visibility into top opportunities, key interactions, or highest risks associated with a sales territory, service responsibility, or outcomes linked to go-to-market efforts. Over time, these models improve to increase the accuracy and contextualization of results provided to users.
Einstein Bots allow Salesforce administrators to build natural language processing-enabled bots through a click-based interface to support high-level interactions and frequently-asked questions with customers. In addition, these bots will be designed to support handoffs to live agents when the bot’s ability to support clients has been exhausted.
Although Salesforce Einstein is already generally available, myEinstein services are still generally in pilot and scheduled for 2018 releases. Einstein Prediction Builder and Einstein Bots are expected to be generally available in Summer 2018. Einstein Vision for Image Recognition is generally available as of now. Einstein Vision for Object Detection is currently in beta and is expected to be generally available in early 2018. Einstein Language for Intent and Sentiment is expected to be generally available in early 2018.
Amalgam Insight’s Point of View: AI on AI
One of the core challenges in enabling machine learning and artificial intelligence is in creating the data models that serve as the core logic for AI. From a technical and mathematical perspective, the combination of statistical and programmatic skills needed to support accurate regression and algorithmic analysis between data sources and a proposed outcome can be difficult to procure, especially for small and medium businesses who may not be ready to hire a data scientist. At the same time, correlation is not causation. Just because a model is mathematically correct does not mean that it is constructed correctly. In fact, a model can be “too accurate” by overfitting variables into a model or by mistakenly adding variables that are outcomes, rather than drivers, of success. For instance, noting that “Closed-Won” opportunities are more profitable than “Closed-Lost” opportunities is statistically accurate. But it doesn’t actually help to understand the sales process. Without understanding what is being analyzed, it can be easy for data modelers to create an extremely accurate model that is both pointless and worthless.
Salesforce has taken an important step forward in opening up Salesforce data to artificial intelligence efforts in a way that allows administrators and operations managers to effectively deploy AI. By providing a combination of statistical tools, AI APIs, and human judgment inputs into the AI development process, Amalgam expects that myEinstein will allow Salesforce users in 2018 to finally understand why Einstein is an important step forward for Salesforce. With this development, Salesforce has truly embraced a core belief of Amalgam Insights in that “AI is the New UI” for allowing end users to quickly access the information that they need.
By 2020, any enterprise application that has not deeply embedded AI for bots, sentiment analysis, and risk analysis will be considered a legacy solution ripe for replacement. Salesforce is moving ahead of its competitors across the enterprise application spectrum by enabling all Salesforce data with both Einstein Prediction Builder and Bots. A key to Salesforce’s success in launching myEinstein will be in uptraining all admins, sales and marketing operations professionals, and developers who will be involved in maximizing the value of Salesforce data through the Trailhead program.
Recommendations
Compare Einstein Bots to existing bot solutions. Companies such as Neuraflash, Troops, JustBots, and the appropriately-named Sudo have already developed Salesforce-enabling bots. Salesforce has not announced a Trailhead training path for Bots as of yet. Depending on the functionality launched, these Bots may either be a replacement for existing bots or serve as a rapid development platform to augment general-use bots that work across multiple platforms. But be aware that enterprise bots are still an emerging category requiring a close eye on feature functionality where no two offerings are alike or fully comparable to each other.
Salesforce administrators should start with understanding the Prediction Builder function. Amalgam believes that myEinstein will quickly become the default interface for translating Salesforce data into both predictive and bot-based outputs. Salesforce customers that have not invested in AI as of yet should find myEinstein to be a useful and design-friendly starting point. Although Amalgam believes that Salesforce administrators should be aware of the full Einstein initiative, we especially recommend that Salesforce administrators start with understanding the Prediction Builder function as this capability provides the most direct translation of Salesforce data into predicting a tangible outcome.
In providing this capability, Salesforce provides a Salesforce-specific solution for AI model building. Machine learning workbenches and automated model building/training were among the most popular functionalities that Amalgam saw at the Strata Data event in New York City. This capability is core to the ability to quickly launch AI.
In closing, myEinstein is an important step forward for enterprise AI and combining the predictive modeling needed for AI with practical human judgment in defining the right data that should be prioritized for Ai. Amalgam believes that the go-live date of myEinstein will be the most important date for Enterprise AI in 2018 as it represents the day that AI will become practical and available to a broad business audience across industries, verticals, company sizes, and geographies.