Salesforce Inc. today announced major updates to its Data Cloud, part of the company’s Einstein 1 Platform, which will allow companies to ingest unstructured data such as PDFs, emails, audio and social media content and connect them with artificial intelligence-powered experiences.

All of these improvements are aimed at greatly improving Einstein Copilot, Salesforce’s generative AI assistant, currently in beta mode, so it can use an organization’s structured and unstructured data.

Data Cloud acts as a pool of a company’s metadata-enabled objects in one place, allowing all the information that an enterprise has to be accessed by services and products, including Einstein Copilot, using a hyperscale engine. This reduces the amount of data fragmentation across enterprise data stacks between cloud, social and mobile and provides access to information all in one place. Now that Data Cloud has access to unstructured data as well as structured data, enterprise customers have reach that they didn’t before.

That matters because real-time relevant data is crucial for producing compelling customer experience and driving accurate AI answers and insightful analysis for customer relationship management systems. However, most information about what’s happening in the real world is tied up in unstructured formats, Rahul Auradkar, executive vice president and general manager of unified data services and Einstein, told SiliconANGLE in an interview.

“Even though data is essentially the foundation for having accurate and effective engagement through CRM and AI, 90% of enterprise data is unstructured data,” said Auradkar. “And this could be emails, it could be PDFs, it could be notes that you wrote to yourself, audio transcripts, video, transcripts, social media posts. But it can’t be effectively used.”

This new capability is made possible with a new vector database support that enhances Salesforce’s ability to handle diverse data types. Now both structured and unstructured data are automatically converted into formats across the Einstein 1 Platform through Data Cloud that can be used for analysis, greater comprehension, insights and more.

“In the past, there was no there were no mature large language models — now we have mature LLMs,” said Auradkar. “So we can take that unstructured data that refers to all the knowledge articles and all the interactions and send it to the LLM. I can now reason over data that is both structured and unstructured — and ask questions of it that has meaning behind it, where before I needed to carefully structure my queries.”

As a result of making these data types available in Data Cloud, and further to Copilot, AI agents can be built that understand what’s happening from knowledge sources better than ever before. For example, a self-service Einstein Copilot-powered chatbot for customers could pull relevant details out of unstructured sources to answer questions by using information from audio and articles linked to customer data that wasn’t available before.

Marketers could take advantage of unstructured data from survey data to build better campaigns and then use that to have an AI assistant iterate their email mailer templates to better target their audiences. Preparing for sales meetings could be done more swiftly by having the Copilot listen to the audio or read unstructured data from a customer’s 10-K or past email interactions.

Even information technology and operations teams would benefit greatly from unstructured data fed through Data Cloud, such as machine logs, sensor readings, images and audio recordings, the company says. Using Tableau, Salesforce’s interactive data visualization tool, the data could be analyzed and Einstein Copilot could provide insights into anomalies and statistical variances in the data points, using semantic similarity that would help engineers discover problems with software or equipment.

Image: Salesforce