Salesforce users are awaiting the rollout of a series of products previewed at Dreamforce last year. This includes Einstein GPT and the extensive cost of Copilot-generated AI tools, AI Cloud, Einstein 1 and Data Cloud platforms. Some are released, others are in beta or in development.
In this interview, Rahul Auradkar, executive vice president and general manager of both Data Cloud and Einstein, talks about the interwoven nature of AI, data, and security in the enterprise and how Salesforce is building its platform with these capabilities in mind. Talk about how you're building it.
Editor's note: This Q&A has been edited for clarity and brevity.
You're responsible for both Data Cloud and Einstein. Explain how data and AI are so closely related that Salesforce built an org chart around this idea.
Rahul Auradkar: We looked at data-driven and AI-driven customer experiences as two sides of the same coin to build a data cloud. Einstein — what we do with predictive AI — has been around since 2016. We know he makes over 1 trillion predictions per week. That number is only increasing significantly every year. On the generative AI side, we've been shipping Einstein GPT products. We talked about co-pilots in Dreamforce. [2023]will be ready soon.
Einstein 1 combines our data platform that has made us successful with Data Cloud and everything that comes with it. That's why Data Cloud is an integral part of the Einstein 1 platform. We provide AI use cases and data-driven use cases through our CRM services. Our partners are doing it, and so are our customers.
“Customer 360” has meant many things over the years: a philosophy, a set of products, and more. What does Customer 360 mean today?
Auradkar: What that means is that our customers can become our customer companies. They provide customers with a fully integrated and consistent experience across sales, marketing, service, and commerce. Whether you want to deliver analytics through Tableau or data integration through something like MuleSoft, the idea is to provide a very rich integration experience that is consistent across all touchpoints, modalities, and channels.
A good example is a user who provides a service to a customer. Customers will definitely appreciate the fact that the company knows their customers well. For example, I opened some tickets, you've already sent me a lot of marketing emails, I bought these 7 of his products from you, and these 3 of his I'm having trouble with one of my products. My profile is a very personal profile and this is a very personal view on the issues I have. This is achieved using a combination of data and data clouds to bring structured and unstructured engagement data to life.
please explain what you mean vector database This is for those who don't understand why Salesforce line-of-business users are important to AI.
Auradkar: We continue to have unstructured data such as Word documents and PowerPoint documents, all on hard drives or in the cloud. More than 80% of his corporate customers' data is unstructured data, such as audio transcripts, call transcripts, and documents related to contracts. Until now, the only way people could actually bring unstructured data to life was through text searches for similarities, or finding documents.
With the advent of LLM [we have] Embedding, a format for models within an LLM landscape. [This enables AI to] Infer unstructured data in a more semantic way and break unstructured data into chunks. These chunks are the vectors we are interested in.
[Think] A story about cats, dogs, and bears. All three are fluffy animals. Therefore, he has one way to observe them. They are all vague. But when you say “fluffy domesticated animals,” you move away from bears and closer to cats and dogs. If you approach a “domesticated animal with claws”, move away from the dog and approach the cat. These are all dimensions that allow us to approach its semantic meaning. They are represented as vectors.
At the World Economic Forum meeting in Davos earlier this year, Salesforce CEO Marc Benioff said: AI can lie and hallucinate, but it can also be a useful technology. Is it difficult to get people to trust AI, even if Salesforce does its best to build in transparency and security features?
Auradkar: So you're asking about hallucinatory works. Predictive AI has a probabilistic approach.Is there a tendency for something to happen? Is there a threshold for when something happens? [negative or positive] It's expected to happen between customers, so something needs to be done about it. You need to contact or send an alert to the person managing that customer.
About generative [AI] On the other hand, there is no definitive response we get. For example, you might ask a generative model to create a response to send to that customer. If you run it again, you might get a completely different result. Because it's not conclusive. So it gets even more difficult, but it all depends on what you ask and how you ask. That's important. Context is important, and how you ground it in the way you ask questions matters. This is where data, AI, CRM, and even trust come into play. Are you using the right data in a reliable way to justify the question you're asking?
What can you tell us about what we can expect later this year regarding the Data Cloud and Einstein roadmap?
Auradkar: Data Cloud is our hyperscale data platform. We use this to deliver a consistent, integrated experience across sales, service, commerce, and marketing. We bring analytics to life, and we bring AI to life. In addition to this, the biggest thing we've done is breathe life into what we call “data locked inside the enterprise.”
We are an open and extensible platform. We're not creating data silos. We are not creating the concept of data gravity and data silos in the sense of saying that if your data is in Snowflake, Databricks, Azure, or Redshift, it will coexist with them. His ETL and copy from them will be zero and he will make it happen in the flow of work. We've shipped integrations with Snowflake, we've shipped integrations with Databricks, and we're doing the same thing with Redshift. Several other vendors do the same. We employ open source standards from the storage layer to the semantic protocol layer. We want to expand that roadmap to more vendors and more integrations.
That's one of the big areas of our roadmap. The second big area is piloting and implementing the unstructured data that I talked about earlier.
The third area of major investment in the data cloud is an integral part of the Einstein 1 platform. To that end, Data Cloud will be deeply integrated into our core platform and begin to illuminate further service and sales use cases. [and] marketing. That's a lot of work.
AI regulation will likely come from the US, UK and EU. How do you prepare for this in product planning?
Auradkar: The good news is that Salesforce has an Office of Ethical and Humane Use. To that end, we also have an executive vice president. Paula Goldman is vice president.she [tracks] all regulations and [represents Salesforce in conversations with regulators] Monitor whether AI is used ethically and humanely.
We look forward to continued discussions at the public policy level. And we will have a strong voice in public policy discussions about what AI actually delivers and what it means. Because we start with trust as our foundational principle, we are probably far ahead of any public policy in providing trust to our customers. Whatever policies come out, we will have an impact on them. And we will be the first to actually step forward and say, “That's right.”
Don Fluckinger is a senior news writer for TechTarget Editorial. He is responsible for customer experience, digital experience management, and end-user computing. Any tips? Please email him.