Consumerization of AI: The Case of Salesforce Einstein

The pace of change in an industry where Artificial Intelligence (AI) technologies are involved is nothing short of astounding. We are seeing projects moving from initial discussions to proof of concept (PoC) and beyond, literally in a few months’ time. Thus, not surprisingly, boards are paranoid about automation and AI, but they are struggling to turn this into actionable mandates. This paranoia stems from many reasons, but two clusters of issues stand out:

1. The fear of business model disruption by leveraging those innovations.
2. The concerns about the investments into technology and talent necessary to stay competitive.

HfS had the opportunity to discuss these issues with executives of Salesforce around their Einstein AI platform. The value proposition of Einstein is focusing exactly on those two clusters, namely to accelerate the time to value for products and services and to contain the costs for infrastructure and data scientists to run AI projects.

Einstein Is an Extension of Salesforce’s Platform Strategy, Not a Standalone New Offering

Even though HfS believes that AI will shift the balance of power to the Mega ISVs, Salesforce is pushing the capabilities of Einstein not as a standalone offering as we have seen with the likes of Google, Microsoft, and AWS, but as an integral part and foundational layers of its platform approach around Lightning. Thus, Salesforce positions Einstein not as a new cloud platform but as a new layer that integrates with existing offerings. Typical use cases or questions, more precisely, are: Which lead should be converted? and Which case should be escalated? A key element in those use cases is the ability to write back to the platform and thus close the loop. Below are examples how Einstein capabilities are enhancing the existing offerings.

• Sales Cloud Einstein: Analyses CRM and activity (e.g. calendar, email) data with a focus on lead scoring, opportunity and account insights, activity capture, and automated contacts. In the next release this will get extended to sales forecasting and opportunity scoring.
• Marketing Cloud Einstein: Provides marketers with journey insights, segmentation, social insights, and recommendations.
• Analytics Cloud Einstein: Leverages discovery by identifying patterns insights in data, be it Salesforce or otherwise, embedding insights directly into business workflows and apps, making analytics easier and more intuitive, collaborative, actionable, and accessible on any device.
• Commerce Cloud Einstein: Offers personalized shopping experiences with commerce insights, recommendations, and predictive sort.
• Community Cloud Einstein: Offers collaboration between customers, partners, and employees with recommendations, feed insights, and search.
• Einstein Platform Services: Offers developers ability to develop apps leveraging natural language processing and image recognition, in particular, sentiment, intent, object detection, and image classification.

Einstein Goes Beyond Low-Level Machine Learning and Next-Best-Action Activities

Similar to AWS, which talks about “democratizing AI”, Salesforce uses phrases like “everyone’s data scientist” and “democratizing deep learning”. While one can easily debate the value of such statements, the commonality lies in the intent to provide AI capabilities out-of-the-box without the need for coding or highly specialized knowledge. Another way of thinking about it is putting standard intelligence into the core products. In the case of Salesforce, these ambitions go one step further by suggesting clients want “clicks not code”. Yet, these ambitions should not be mistaken for assuming Einstein is offering only low-level machine learning around recommendations and next-best-action activities.

To get an understanding of the depth and scale of Salesforce AI ambitions as well as capabilities, one only has to look at its M&A activity. Salesforce has spent north of $1 billion on assets that include AI related capabilities over the last three years. Exhibit 1 gives an overview of this activity. Of these acquisitions, Krux alone required $800 million while RelateIQ cost $390 million and BeyondCore another $110 million. Those capabilities also underpin the Enterprise Platform Services, which include sentiment and intent analysis as well as object detection and image classification, thus taking Salesforce far beyond low-level machine learning and next-best-action activities. Similarly, pointers for the roadmap that include case classification, chatbots, broad-scale natural language processing, as well as integration with IoT events show that Salesforce AI ambitions go much deeper than most other cloud platform plays.

Exhibit 1: Salesforce’s M&A investments in AI

Source: HfS Research, Salesforce, 2018

In our view, the most tangible asset that references both Salesforce’ investments and its ambitions in AI is the Einstein Vision API. By leveraging Computer Vision, this offer integrates image recognition into the core CRM offerings as well as into custom apps. The use cases outlined below highlight the complexity that Salesforce is aiming to provide.

• Visual search provides customers with visual filters to find products that best match their preferences and takes photos of products to discover where customers can make purchases in-store or online.
• Brand detection allows marketers to increase marketing reach, brand integrity, and ad campaign ROI by analysing user-generated images in communities, review boards, and social media. This allows building a better understanding of customer preference and lifestyle to improve customer experiences.
• Product identification streamlines sales and service by giving reps the ability to evaluate product issues before sending technicians into the field, discover which products are out of stock or misplaced, and measure shelf-share to optimize product mix and selling potential.

AI Strategies Have to Focus on Use Cases

If you are a hospital looking to automate diagnosis or patient management or if you’re a bank looking to optimize thresholds for anti-money-laundering (AML) activities, Einstein won’t be of too much help. But for customers looking to optimize CRM, sales, sales, service and marketing processes, Einstein is a decidedly welcome addition to their quiver of transformation levers. As with the broader notion of Intelligent Automation, organizations need to focus on evaluating which processes are conducive to significant enhancement by AI. To find your way through the AI maze, HfS has suggested clustering the different starting points for AI projects, namely the extension of RPA, conversational services, autonomics, the expansion of data science, and machine learning. Exhibit 2 highlights the different starting points for AI projects and the investment requirements.

Exhibit 2: The Journey Toward AI Has Disparate Starting Points

Source: HfS Research, 2018

From a broader service delivery perspective, one could add industrialization, data-centric alignment, and domain-specific offerings as additional segments. In our view, those segmentations can help to provide more clarity for the context of AI discussions and therefore more relevance when comparing approaches. Einstein’s sweet spot is around industrialization and machine learning. Or, put another way, it is aiming to make machine learning both seamless and ubiquitous. This relates predominantly to specific CRM activities, not horizontal enterprise applicability. At the same time, organizations have to be cognizant of the fundamentally different quality of Einstein in comparison to AI frameworks or cognitive libraries from the likes of Google, Microsoft, and AWS. In summary, Einstein is not a foundational AI layer across the enterprise, but it helps to significantly accelerate the time to value for CRM and sales related activities.

Bottom Line: AI platforms are an increasingly important lever to transform service delivery, but only as part of a broader service orchestration.

Salesforce Einstein is a telling example that in order to industrialize machine learning, deep learning and natural language processing (NLP) and, over time, broader AI capabilities, providers need deep investments. That is, if they want to help organizations to advance on their journey toward the OneOffice and autonomous processes and not just emphasize recommendations or next-best-action activities. Salesforce’ domain-specific focus is a clear way of differentiation, but organizations evaluating those capabilities need to do this in the context of the broader digital and Intelligent Automation strategies. It all boils down to managing data and automation initiatives with the notion of service orchestration in mind. Only then will organizations progress to scaling out their projects.

In a noisy market where AI is neither defined nor do organizations fully understand the divergence of use cases ranging from low-level recommendations to highly complex data science projects, Salesforce needs to demonstrate the learnings from the early deployments to reference the maturity of its approach. It also needs to articulate more clearly how its partner ecosystem can leverage those much broader capabilities being integrated into the platform. At the same time, discussing AI in the context of Salesforce references how fast the company has grown up and how far it has come from its beginning as no-frills CRM SaaS platform. AI will both broaden its capabilities and blur the boundaries of its positioning.