When Salesforce announced its spring release this week, it revealed that its artificial intelligence platform, dubbed Einstein, can build data models automatically, even when customers have customized their products to meet the company’s unique requirements.
They called this “using artificial intelligence to generate artificial intelligence,” and it’s not something they have talked about before when describing the platform’s capabilities.
John Ball, GM of Einstein at Salesforce, said they did this for a couple of reasons. They wanted to make this process of integrating intelligence as easy as possible for customers, but from a data standpoint, they also knew the more data the customer provides and the more custom data fields provide new signals, the more robust the models that it could generate from that data.
“When you are customizing and extending Salesforce, you are doing it for a reason because that models your business process,” Ball explained. “Einstein can look at the underlying metadata and figure out the schema and relationships to know this is email or a phone number [or whatever it is] and include that in the data set. It may or may not be predictive, [depending on the data type], but you can throw the kitchen sink at it and we figure it out using math,” he said.
They take this a step further though. As the intelligence layer generates these automated models, Ball says the underlying software continually tests them and pits them against each other in a competition of sorts. It picks the most robust models and discards the others.
Salesforce has been slowly integrating Einstein since it was first announced last fall, trying to integrate AI into every piece of its vast software platform — with varying degrees of success. With today’s announcement of the spring release, it appears that it touches just about everything now, at least to some extent.
Simply having tech for tech’s sake doesn’t do much good, of course, and Salesforce still needs to prove that the AI-fueled capabilities are helping customers improve their experience over time.
Disrupt 2026: The tech ecosystem, all in one room
Your next round. Your next hire. Your next breakout opportunity. Find it at TechCrunch Disrupt 2026, where 10,000+ founders, investors, and tech leaders gather for three days of 250+ tactical sessions, powerful introductions, and market-defining innovation. Register now to save up to $400.
Save up to $300 or 30% to TechCrunch Founder Summit
1,000+ founders and investors come together at TechCrunch Founder Summit 2026 for a full day focused on growth, execution, and real-world scaling. Learn from founders and investors who have shaped the industry. Connect with peers navigating similar growth stages. Walk away with tactics you can apply immediately
Offer ends March 13.
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.
It’s worth noting that HubSpot released a research report in January that found 63 percent of respondents don’t realize they are using AI technology even when it’s there, and really why should they, any more than most users understand any algorithm that’s running the software they use.
Ball acknowledges that end users aren’t thinking about any of this when they use Salesforce products, but says buyers certainly are hearing about AI technology and are asking about it in sales calls.
Even though Salesforce and IBM, which announced a partnership yesterday afternoon, have insisted on humanizing their AI platforms by calling them Einstein and Watson, 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.
