Is there any promise for the use of AI in sales and marketing? In a B2B context? Leading CRM solution provider Salesforce seems to think there is. In the past year, they rolled out AI-enabled enhancements to their cloud-based sales, marketing, and support solutions that are designed to deliver more predictive analysis, helping sales reps identify the most qualified leads, and giving marketers the intel to know who to target with what offer.
To determine whether there is hope for such solutions or just hype, we’ll take a quick look at the major features of Salesforce’s Einstein AI, review some of the early critiques by the experts, and ponder some of the real-world use cases that might yield breakthrough results. Salesforce has deployed Einstein across their entire suite of solutions, but for the brevity’s sake, we’ll focus just on the sales cloud.
Feature overview of Einstein Sales
- Einstein Lead Scoring: Einstein Lead Scoring models are built specifically for each customer and organization, which ensures that the models are tailored to the business. Einstein Lead Scoring analyzes all standard and custom fields attached to the Lead object, then tries different predictive models like Logistic Regression, Random Forests, and Naïve Bayes. It automatically selects the best one based on a sample dataset.
- Einstein Opportunity & Account Insights: Sales Cloud Einstein analyzes all the standard fields attached to the Opportunity data in addition to email and calendar data, and then uses machine learning, natural language processing, and statistical analysis to provide sales reps and managers with "Predictions", "Key Moments", and "Smart Follow-Ups."
- Einstein Activity Capture: This logs historical emails and calendar events from up to six months back for Gmail and up to two years back for Office 365. It then works in the background to passively capture every email or calendar event sent or received. The captured emails and events are all displayed in an activity timeline, providing a history of the team’s relationship with a customer.
- Einstein Follow-Ups: This provides proactive email notifications, letting reps know when a customer needs an immediate response, or set a follow-up reminder.
Having lived through many “hype-cycles” over the years, technology buyers tend to react in the same way whenever there is some breakthrough new technology: “so, what problem does it actually solve.” In a recent article on new AI solutions, NextWeb talked about how “AI-powered tools are now helping scale the efforts of sales teams by gleaning useful patterns from data, finding successful courses of action, and taking care of the bulk of the work in addressing customer needs and grievances.” Techcrunch takes a bit more pragmatic view on Salesforce's AI, “certainly automatic model generation, if it works as described and truly delivers the best models in an automated fashion, is highly sophisticated technology, but in the end, users don’t care about any of that. They want tools that help them do their jobs better, and if AI contributes to that, all the better.” On how to think about AI in the technology solution stack, they noted “the fact is AI is not a product in the true sense, so much as a set of technologies. We should keep that in mind as we judge these announcements, looking at how they improve the overall products and not at the shiny bells and whistles.”
Possible Use Cases
Complex B2B sales remains a mostly human activity, and any technology deployed to support the process should help augment, not replace human judgment. If applied correctly, AI could help spot consistent patterns that narrow down a list of highly qualified leads for reps to contact given certain triggers. This is no doubt useful and could drive efficiency, but if the objective is to close larger more complicated enterprise sales, the most likely use case could be AI that tells reps who to talk to, but not what to say or do next. As we have discussed before, buyers and the buying process is not perfectly rational, and algorithms need good data.
CRM systems can be full of human-keyed data that may be inconsistent, inaccurate, or lack sufficient depth to be meaningful. Additionally, much of what’s entered can be subjective (close dates, probability of close, deal size) and often overly optimistic. What ultimately matters are customer behaviors: what products did they buy, when did they buy, what did they pay. Using actual prior transaction data for the AI analysis would likely improve relevancy and accuracy of predictions to make marketing and sales more efficient, and more importantly, more productive.