Looking for something else?
Artificial intelligence has become a battleground for major technology providers. But some vendors are trying to leapfrog their competition by making AI more practical, bringing intelligence to workers' daily tasks.
Salesforce is one such vendor intent on improving workers' day-to-day activities with its flavor of artificial intelligence (AI). During the past year, through acquisitions and internal development, Salesforce AI has arrived in the form of Salesforce Einstein. Salesforce, the cloud-based CRM vendor, will integrate Einstein into various clouds in its Salesforce Customer Success Platform, from the Sales and Marketing clouds to the new customer experience offering Commerce Cloud. By building AI natively into workers' existing tasks and using Salesforce data, the company believes it has an edge over competitors.
AI can enhance capabilities like predictive lead scoring (software that assigns numerical values to sales leads to identify the most promising), marketing campaign management and customer service, said Salesforce chief scientist Richard Socher in a conversation with SearchSalesforce about Einstein.
"Where can you use AI to enhance someone's workflow?" Socher said. "How can you continuously collect data so that AI elements get smarter and smarter, and then surface them back to people so they are empowered?"
Socher outlined the kinds of efficiencies workers can derive with Salesforce AI. He explored the example of AI in common sales processes. AI can surface data in email and calendaring to make recommendations to sales reps and help optimize their time.
"We can go through email and help you understand your calendar, your schedule and help make smart recommendations on what to follow up on and whom to talk to," Socher said. "No one wants to make a sales call that isn't wanted."
While Socher acknowledged that technology will have sweeping effects on human work, he remained steadfast that AI will bring new job opportunities. "We aren't creating self-driving cars. We created new jobs based on new kinds of technology." Socher said that ultimately, AI's ability to displace jobs "depends on the use case" and added that it was important not to build biases into the algorithms. Bullish on the future of Salesforce AI, Socher provided his take on Einstein and how AI will take shape in the Salesforce roadmap.
Many major vendors are trying to make their mark in AI. How does Salesforce Einstein differ from IBM's Watson, Microsoft AI or even Oracle's Adaptive Intelligent Systems?
Richard Socher: We are focused on AI for CRM: sales, service, marketing, IoT [internet of things] and a little bit of healthcare. We don't build a general-purpose AI engine that is abstract. We try to focus on use cases. We first make sure you flip a switch and it works. Then, over time, you can customize it and build your own apps. But our core is always CRM.
There are three elements of AI:
- You have to have access to the data to understand it. We have metadata associated with that data.
- The second is the algorithms. You need access to top talent to develop new things. We have a research group and access to great talent.
- And third is workflow integration. It can mean a lot of different things: In CRM apps, it means how you use AI to enhance someone's workflow. How can you continuously collect data so that the AI algorithms get smarter and smarter and then surface that back to people so they are empowered; collecting the data, surfacing AI predictions in a way that empowers and makes users more efficient.
So it's native to the processes that workers are already using?
Socher: We can go through your email as a sales rep that helps you understand your calendar and whom to follow up with. We can surface and score all the opportunities and leads you have. No one wants to make a sales call that isn't wanted or doesn't result in a sale. You can sort by a score of all your leads, and you call those that are most likely to result in a sale.
How does Einstein's predictive scoring or personalization feature extend functionality in Salesforce Sales and Marketing clouds, features that were present prior to Einstein?
Socher: My team is pushing the state of the art in artificial intelligence. That can mean you improve existing features, but it also enables us to build new kinds of products. Now we're working on question answering, where we can use general text and ask generic questions and get good snippets back. Generally, this technology has been in the hands of consumer-oriented companies, and we wanted to bring it to the enterprise customers. It's an ongoing competition on the Stanford question-and-answering data sets; for a while we were No. 1, and now we have some tricks up our sleeves to improve our accuracy. We're taking research ideas. And we publish our research and exchange with the academic community.
Talk about your strategy for rolling this out to various clouds. What are you starting with?
Socher: We're actually working in parallel effort. AI is moving into all the clouds, and it's a parallel process. We have predictive lead scoring for sales, case routing for Service Cloud, working in Marketing Cloud on images to target their audiences. We have various features in IoT. All of them will need AI and are getting it through different efforts.
How might AI affect human work? Could it take human jobs?
Socher: It's important to think through how the technology will affect people. You need to ensure that there are no inherent biases in the algorithm. The answer is quite complex, but in some ways it might empower people.
But if salespeople now aren't wasting time on the wrong calls and are now making sales that work out, a company won't let go of half of its salesforce. In marketing, 15 years ago, we didn't have the social media marketing position, but we created new jobs based on new technology. In some areas there may be less need, but it may empower new jobs and enable people to be successful. There are AI use cases we also don't work on, such as self-driving cars, and that may have a more immediate impact on jobs. Whatever you are working on actively you should think about.
There have been some examples of AI -- such as a Microsoft chatbot rolled out then removed last year -- that haven't been successful.
Socher: We take it seriously -- the trust other companies have in us. We're cautious about AI. If Salesforce worked on a chatbot -- now, this is hypothetical -- the user would have an incentive to work together with the AI chatbot. The feature sets we're working on -- we want to make the job easier and better and more efficient with CRM. So there's less of this kind of attack angle or scenario that you described.
What is the business value of Einstein ultimately? Does that change over time?
Socher: More important than the top business value, it's important to note that AI will be in all the different aspects of enterprise software. You want to empower service folks to focus on the hard cases and give them tips on how to give the right answers quickly in a live setting. You want to empower salespeople to spend time in the most efficient way. You want to empower marketers to understand the whole product in the landscape. All three are important.
Microsoft takes AI beyond Office 365
Salesforce pitches artificial intelligence for all
A look at Enterprise AI and other 2017 tech trends