One of the biggest challenges when building AI apps is how to reconcile the fact that the data scientists who create the AI models and the customers who use the AI models live in two different worlds. Experts may spend a lot of time optimizing new predictions that may not provide any practical value to end users.
Salesforce Einstein Discovery bridges that gap by helping business users create and experiment with AI applications on their own.
The resulting predictions may not be as accurate or scalable as those honed by data scientists, but they can help users identify which models to actually use. Once users identify a good fit, data scientists can feed the resulting algorithms for further improvement.
At the Dreamforce 2018 conference, Einstein Discovery pioneers shared some of the lessons they learned last year.
understand the business case
Auction Nation, a South African junk car trader, quickly implemented Salesforce Einstein Discovery to improve its process for evaluating the prospects of wrecked vehicles. The company employs a group of automotive experts who buy cars from insurance companies to be resold at a markup. Unfortunately, some types of damage are much more difficult to repair and resell for a profit.
Before becoming a data scientist, you need to understand your core business. They need to realize that there is an underlying process to improve.
errol levinAuction Nation COO
Auction Nation COO Errol Levin hopes Einstein Vision will help buyers better identify these lost causes and prevent them from the company's inventory.
Most of Levin's staff believed it was impossible to create an app that performed better than humans, but using Salesforce Einstein Discovery, Levin experimented with his own workflows. . The initial results weren't perfect, but the workflow was improved.
For Levin, it was important to explore and discover these limits for himself.
“You have to understand the core business before you get to data scientist,” Levin says. “They need to understand that there is an underlying process to improve.”
Einstein Discovery's AI-based analytics helps you discover patterns in your customer dataset and identify potential sales based on past performance and how other customers move through the customer journey.
start with a small team
Rebecca Greenberg, Director of Commercial Systems and Specialty Analytics at Takeda Pharmaceutical Company, wanted to provide actionable information to support sales representatives who interact with physicians. However, due to Health Insurance Portability and Accountability Act regulations, much of this data is protected and cannot be shared across the company.
Greenberg experimented with Salesforce Einstein Discovery to explore large amounts of previously underutilized data. She quickly discovered that doctors who ordered Takeda had a bell curve in cancellation rates. This discovery helped identify doctors who were on the verge of cancellation, allowing sales reps to focus more time on contacting these doctors. The resulting app alerts field personnel about accounts at risk of loss while protecting patient data.
For Greenberg, it was important to experiment with what kind of predictions he could make with a small team so he could decide what to focus on.
“That small team allowed us to see what parts of the tool were helpful and where we needed to work on them with real data scientists,” Greenberg said.
Clean up your data
AI predictions are only as useful as the data that powers them. If there's trash in there, trash will come out. Managers who use Einstein Discovery directly to experiment with new models can quickly identify where business processes need to change to improve data quality.
“This will get rid of the trash quickly,” Greenberg said.
Takeda has already made significant investments in understanding, cleaning and organizing data, using tools such as master data management and working with a team of data stewards. However, the process of building new AI models helped identify unexpected issues.
Jonathan Ray, former director of product management for Salesforce Einstein Discovery, said another Einstein user discovered that the company's Chinese subsidiary was using a completely different accounting model than the rest of the company. This contradiction first became apparent when the Einstein app showed sales in China were several times higher than in the rest of the world, even though management intuitively knew otherwise. became.
Once you've identified a useful business case, you may also consider reaching out to a partner to help scale up your new AI app. After Greenberg's initial success, she worked with her LiquidHub, a customer engagement company recently acquired by Capgemini, to scale up the app. She said people are willing to help, sometimes at little or no cost, to build a resume of AI success stories.
Understand how the tool works
TouchCR, a consumer marketing analytics platform, turned to Einstein Discovery to improve the sales process for its salespeople.
Richie Hale, TouchCR's chief innovation officer, said he personally participated in the early experiments to better understand how the business could improve. Once we had some insight into what worked, we were able to provide even greater insight to the rest of our team to expand our use of AI.
His biggest insight is that AI is more of a journey than a destination, and expect many failed ideas along the way. Without experimentation, companies may lose business to more agile competitors.
“If you want to be competitive, you need to have market insights that you don't currently have,” Hale said.
