AI for Everyone! This week we continue our erratic series of under represented companies in AI; this week is Salesforce.

AI for Everyone!

Some companies claim they offer AI for everyone. Just use our wonderfully easy API, connect to our cloud, and you’ll be going in no time. But what if you don’t know how to code? And what if you have a lot of data and want to take advantage of AI/ML but don’t know how to code?

Enter Salesforce

Two years ago, Salesforce realized they needed to bring AI/ML into their products to stay competitive. Salesforce had an interesting choice to make: do they hire data scientists and offer consulting services to companies to make sense of their data or do they just build an AI/ML system that anyone can use?

Salesforce chose to build an automated data science platform, realizing they would never be able to hire enough data scientists to build custom solutions for their clients. However, doing data science the right way is hard; even creating a simple model to predict which one of your sales leads will convert is very hard. A lot of the work a data scientist does is understanding the data, building custom “features” (or variables) that help lead to a better prediction and making sure all of the data is sound. Salesforce would need to package all of this software together and convince data scientists they weren’t taking their jobs. A strange ask.

After reading a detailed article about the experience in Wired, the data scientists at Salesforce don’t seem to mind. Many of them seem encouraged and excited about bringing AI/ML to everyone and Richard Socher, chief scientist at Salesforce, says in his website he enjoys “making AI easily accessible to everyone.”

Salesforce ultimately launched their product, calling it ”Einstein” in 2016. Users were able to take advantage of features such as lead scoring, text mining and other features and consumers have responded positively, currently calling on Einstein to make over 1 billion predictions a day, up from 200mn predictions a year ago. Salesforce is also finding that Einstein is becoming a selling point, not just a cute AI mascot. In a Jefferies survey this past year, 48% of Salesforce partners actively selling Einstein features, up from 32% a year ago.

Salesforce has enabled almost everyone to use AI/Ml and truly make data based decisions. In a vote of confidence, Marc Beinoff, CEO of Salesforce, says he uses Einstein features regularly in meetings and has this very interesting quote:

“I will literally turn to Einstein in the meeting and say, ‘OK, Einstein, you’ve heard all of this, now what do you think?’ And Einstein will give me the over and under on the quarter and show me where we’re strong and where we’re weak, and sometimes it will point out a specific executive, which it has done in the last three quarters, and said that this executive is somebody who needs specific attention”

Imagine getting work advice from an algorithm?

Enough boring data science, what about AI?

Salesforce’s push into AI has come through a slew of acquisitions, including Tempo AI (smart calendar), MinHash (personal assistant for marketing), PredictioIO (machine learning platform) and MetaMind (AI as a service). The most impactful acquisition is clearly MetaMind, as the CEO of the company, Richard Socher, is now the chief scientist of Salesforce

Socher is the perfect “acquihire”: PhD from Stanford in deep learning, then an adjunct professor at Stanford, finally startup CEO. He certainly has influence on the future of Einstein; when looking at Socher’s published papers and patent filings in 2018, it reveals a significant amount of natural language deep learning based systems, including voice analysis and text mining.

This work was highlighted at the recent Dreamforce, where Salesforce unveiled Einstein for voice. An interesting slide in their presentation shows a user recording notes into Salesforce via voice transcription and then having the material extracted via text mining. The idea is that sales opportunities can be updated via voice, rather than making amendments to the system.

Finally, a marketing detour

Currently, marketers can use Salesforce Einstein to personalize user journeys, build a recommendation engine, engagement scoring, recognize logos, objects, food, and scenes in social images. While that is a lot, they could soon solve the holy grail of marketing…

Multichannel attribution

Salesforce recently acquired a company called Datorama, for a cool $800mn. Datorama is a company that helps marketers pull all of their data together in one place, including everything from programmatic to Facebook data. By combining this data with actual business sales data in Salesforce, Salesforce could build a multichannel attribution model that actually works.

Finally, an answer to that John Wanamaker quote

Facebook Comments