Metadata Integration Enables Generative AI in Salesforce Data Cloud

Many organizations are still building their data platforms. Salesforce Inc. has taken a breakthrough step: By seamlessly incorporating metadata integration, the company transformed its modern data stack into a comprehensive application platform. Einstein 1 platform.

Under the leadership of Muralidhar KrishnaprasadBuilt on the company's foundational metadata framework, the platform ushers in a new era of data leveraging by harmonizing metadata and integrating it with AI and automation, according to John McClellan, executive vice president of engineering at Salesforce.

The Einstein-1 platform and its innovation story Salesforce Data Cloudand what are the AI ​​capabilities in Einstein 1? And what can this unified platform bring to your organization?

According to Krishnaprasad, the company set out to build a platform that empowers all business users — salespeople, service engineers, marketers, analysts — to access, use and act on all data, regardless of where it resides. Salesforce's open and extensible platform not only enables organizations to liberate trapped data, it also equips them with generative AI capabilities to deliver next-level personalized experiences to employees and customers.

“Analytics is really important to understand the business, but we also need to make sure that all of our data and insights are actionable,” says Krishnaprasad. “The confluence of these three — AI, automation and analytics — is our big thesis. Bringing in the metadata layer as well is really important. That's the big secret we've put together.”

Krishnaprasad is George GilbertSenior Analyst at theCUBE Research “The path to intelligent data apps” The podcast series is an ongoing effort by theCUBE to explore the cutting edge of intelligent data applications. We delved into metadata integration, the open API technology behind Einstein 1, the key capabilities of the platform, and how its extensibility and interoperability improves ease of use across a variety of data formats and sources.

Metadata Integration: Diverse technology stacks built to work with virtually any IT environment

The platform is built on Trino, a federated open source query engine, and Spark for data processing, with a rich set of connectors and an open, extensible environment that gives organizations a wide range of options built in. Organizations can freely share data between systems, including warehouses and lake houses. But this is just the beginning.

“We also use something called HyperEngine, which is the in-memory overlap engine that we use in Tableau,” says Krishnaprasad. “When you're doing Tableau or other data exploration, you want sub-second response times. We do all of this automatically in-house.”

The platform supports a wide range of machine learning options. Users can even add their own large language models to the stack. Whether they use the Salesforce Einstein Integration Platform, Databricks, Vertex or SageMaker, they can do so without needing to copy data, Krishnaprasad says.

“You can literally point the tool at it, create a model and score your data there. [that same flexibility] “For an LLM,” he said.

The platform includes three levels of extensibility, which Krishnaprasad noted allows organizations to easily standardize and extend their customer journey models. The platform starts with a basic reference model, such as customer name, and can be customized by adding new fields or creating new models and relationships. At the next level, these models can be leveraged to generate a range of insights, including business intelligence and AI-driven insights. Finally, at the highest level, users can introduce their own capabilities and triggers to act on these insights.

The platform also offers a great deal of flexibility in the integration process: for example, while many systems perform integrations only once, the Einstein 1 platform performs that function continuously and also allows users to create different integration graphs.

“Remember, we're a multi-modal system, which means we're looking at your entire customer journey,” Krishnaprasad says. “Your marketing team might decide that if you're advertising, a 60% match is good enough, whereas if you're providing a service to someone, running a bank, or working in finance, you probably want a very deterministic match. We're empowering our customers to create different graphs that they have complete control over. We provide flexibility at almost every level of the stack, enabling them to create the experience that's right for their business.”

Three Elements Driving Transformed Customer Journeys: AI, Automation, and Analytics

Whether batch, streaming or real-time, the platform foundation incrementally ingests, harmonizes and integrates data. This capability creates a standardized metadata model that creates, extends and enriches a single 360-degree view of what customers are doing with their business.

Achieving this view is a game changer for organizations, Krishnaprasad says, because it liberates siloed data, much of which is stored in unstructured formats. This data — which includes conversations, Word documents, PDFs, emails, audio, video and more — is a gold mine, and the platform allows customers to extract and act on it.

“What we've done with this customer 360-degree model is use that integrated data wherever it is, generate insights based on the data, and make this integrated data and insights available across all application surfaces, and in many cases even take push actions so you can react to these things,” Krishnaprasad said. “That's the larger customer journey we've enabled.”

For example, when a customer sees an ad and then visits a website, a company knows who that customer is. Salespeople know exactly what site visitors are interested in. Service reps understand what customers are unhappy with. Analysts have easy access to all the information they need to generate insights that will strengthen the business. Through these capabilities, customer engagement improves exponentially, according to Krishnaprasad.

“When you combine that with generative AI, you can take that to the next level, and then you can start to self-service a lot of these things,” he said. “You want to make sure you're giving the right answers, and that's what our data platform allows you to do. We're taking that data to a higher level so that we can create unified models and power a unified experience across the entire customer journey.”

Below is the full conversation, part of our “The Road to Intelligent Data Apps” series.

Image: Ralph Hahn/Getty Images