Getting to Know Einstein
Einstein once said, “Anyone who has never made a mistake has never tried anything new.” I have to give Salesforce some credit here. They are always pushing the envelope and trying new things. With that said, they have made their fair share of mistakes as well. That's why while attending Dreamforce this year, I was intrigued but skeptical about the newly announced Salesforce Einstein. Einstein is Salesforce’s way of introducing Artificial Intelligence (AI) onto the platform. The whole thing was quickly glossed over in the opening keynote and left me craving some gritty details. I made my way over to the Einstein demo area afterward in hopes of getting a glimpse behind the scenes. Unfortunately, more flashy demos about how fast “Steve Benioff” can sell a widget to “Mark Jobs” with the help of Salesforce Einstein ensued. Finally, I was able to get some one on one time at a kiosk with a guy who was actually configuring some of the tools behind Einstein. Score.
I know what your thinking. How do you configure AI? Shouldn’t it just tell me what I need to know and finish responding to my emails for me already? Einstein is made up of a few key tools: Machine Learning, Deep Learning, and Natural Language Processing. Let’s break these down a bit.
Machine Learning
- This is the process of looking for patterns and trends in your data and returning recommendations based on predictions according to the trend. This could help predict the reaction of a customer to an email marketing campaign based on the reaction of similar customers reacting to similar campaigns.
Deep Learning
- Deep Learning is most notably used to process images. Facebook uses deep learning to recognize faces in your photos. This allows you to ‘teach’ Salesforce to look for things you may be interested in, such as your branding in the image of a social media post.
Natural Language Processing
- The name actually explains this one pretty well. This is how a computer reads and ‘understands’ text. It can allow the recognition of more complex meaning within a text including sentiment and emotion, offering further insight into the mind of a customer or prospect.
My demo at the kiosk walked through an example of setting up Deep Learning to recognize if an image of a car was an Audi or a BMW. We started by uploading a few thousand images of each one. Based on what it had learned, we gave it a few single images and asked it to identify what it was. We give it an image of a BMW 3 series and Einstein responded with 98% certainty that this was a BMW. Nice one Einstein. There didn’t seem to be anything glaringly missing here. It may take me a while to amass that many images of one type of car, but beyond that, the configuration was fairly straightforward and effective.
AI by definition is the appearance of intelligence exhibited by a machine. So how intelligent is Salesforce Einstein? Only as intelligent as your data. Each of the tools relies on accurate data in large amounts in order to operate effectively. It's unclear what sort of limitations or difficulties small and new organizations may experience while trying to take advantage of certain aspects of Machine Learning.
My takeaway here is that Einstein is not another Siri or Cortana with preset functionality and commands. It’s a fully configurable set of intelligent tools that can be applied directly to business applications on the Salesforce platform. Of course, the future of these tools is much more exciting than the initial offering, but I for one, can’t wait to dive in. As AI turns the corner into Turing test territory we’ll all be out of a job anyway.
What Salesforce is doing with AI is bold, but similar to lightning, it may be a while before Einstein really takes hold with a broad audience. We’ll have to wait and see where this takes us.
How do you feel about Einstein? Want to share your experiences or opinions? Share them with me on the Arkus Facebook page, in the comments below, in the Success Community, or to me directly via Twitter at @jbujold.