Salesforce Exec: ‘Potential is Enormous’ for Machine Learning in CRM

When it comes to machine learning (ML) as it applies to customer relationship management (CRM), we’re still “in the first inning,” but “the potential is enormous,” according to Mike Rosenbaum, EVP of product at Salesforce.

“There’s two things that I would say are helping us drive continued growth in Sales Cloud,” he said July 10 at an investor meeting in San Francisco hosted by RBC Capital Markets.

“Number one is this idea that – and this started a few years ago really – where you don’t want to think about Sales Cloud as just an application,” but something that “really becomes a platform around which you run sales-related business processes at a company,” he said.

The “most obvious example of that is CPQ,” he said, referring to the configure-price-quote functionality that it obtained in late 2015 through the acquisition of SteelBrick. CPQ enables companies to provide customized price quotes to each of their customers.

He added: “I really see a day when 100 percent of Salesforce customers could be using CPQ. And that creates the potential not just to continue to grow Sales Cloud, but I think really to grow the whole category and really just continually think about all the business processes that are related to selling and how Salesforce becomes a platform around which you just more efficiently run those business processes. And so, continuing to add applications to that portfolio that are well-integrated to Sales Cloud is just a super way for us to continue to grow our market share in that category.”

But he said, “the more interesting — the more exciting maybe — I’d say product-related initiative is about” Salesforce Einstein artificial intelligence (AI) and ML as it applies to selling.” Because of Einstein, “I really think that we are on the cusp of a very revolutionary change in how Sales Cloud works,” he said.
Before Einstein, salespeople were always asked to enter sales info for Salesforce to gain insights and help companies sell more effectively, he noted. But the addition of ML enhances the quality and usefulness of the data and “will become the primary value that an organization gets from Salesforce automation,” he predicted.

He added that, “for someone like me, [it’s] somewhat magical when you see a machine learning-derived forecast automatically applied to a company’s Sales Cloud implementation, and we are now automatically creating models on behalf of customers that score each and every opportunity and give sales managers and salespeople insights that they just didn’t have before” Einstein.

“The technical breakthrough here is that the machine is developing the algorithm that’s used to do the forecast, or that’s used to do the score, or that’s used to predict the right leads to follow up with – that it’s not a person,” he said.

Salesforce has “democratized” AI and ML for its customer base, he went on to say. After all, he explained: “Every company in the world can not afford to hire a team of data scientists that have the skills, the expertise and the experience to build machine learning models on top of the way they use Salesforce. Some companies can do that — and we encourage them to and we work very closely with them to help them do that and we love that…. But each and every one of our customers can’t afford to do that.”

And that, he said, is “why Einstein is such a powerful idea.” Salesforce “can do what the top companies in the world can afford to do [but] we can do it for everyone. And we can do it automatically and we can do it efficiently,” he noted.

Asked if there are any hurdles to customers adopting Einstein, he said: “To some extent, there is a small amount of work, I think…. A customer has to use Salesforce in a way that aligns well with how we built Einstein to ingest the data. It gets back to the question of the more data that you’re entering into the system, the better the models can be that we generate. So, for something simple like lead scoring, we can score leads but we need the customer to have at least a thousand leads in the system with outcomes that we can use to build the model … . So, I think that’s partially like a little bit of a business process evolution that needs to take place.”