By Stuart Lauchlan

December 15, 2023

The Data Cloud has rightly been one of Salesforce’s biggest product focuses this year. As the company’s World Tour series of events hit New York yesterday, the latest news was that it will soon be able to handle unstructured data via the Data Cloud Vector Database. 

This provides integrated vector database support that means that unstructured data, such as emails, audio, social content etc, can be combined with structured data, such as purchase histories or product inventories. 

As per the official announcement, a use case exemplar might be: 

For example, customer service leaders will enhance efficiency and customer satisfaction by utilizing a platform that proactively presents relevant knowledge articles to service agents the moment a case is created. This allows for quick identification of similar cases and the integration of automation, thereby reducing case resolution time and improving the overall customer experience.

The Einstein Copilot generative AI assistant will also be able to tap into unstructured data. Also announced in New York was Einstein Copilot Search, which will use structured and unstructured data from the Data Cloud to provide suggestions for searches. As a use case exemplar, Salesforce pitches that it would be able to link a customer’s emails and phone calls to their formal support ticket history to assist a customer service representative.

Data, data everywhere

According to stats from Forrester Research, some 90% of enterprise data exists in unstructured form, rendering it inaccessible for business applications and AI models. The research firm also predicts that the volume of such unstructured data will double next year. 

Getting on top of the data problem is crucial, said Salesforce co-founder and Chief Technology Officer Parker Harris:

Your data is disconnected. Your APIs are disconnected. You've got a bunch of different apps, they're not connected. You've got vendors who don't work with each other that aren't connected. You've got low productivity. That is not what you want. That's not the AI experience that you want.

The Salesforce solution to this is the Einstein 1 platform, first announced back in September at Dreamforce in San Francisco. Harris explained: 

It’s integrated. It's one platform. You might think of [Salesforce] as multiple platforms. We've done a lot of acquisitions over the world. One of my main focuses is to make sure we bring all that together, that it's integrated into one platform, that it's AI-first, it's automated, that it's low code, no code all the way to pro code. It's everything. And that it is completely open.  

So whether you're using Sales or Service or Marketing or Commerce, whatever application you're using, or you build applications on our platform, it's all powered by metadata. You've got the Data Cloud, not as some separate data cloud, not off to the side where you're like, ‘I got all my data here, come and get it’. No, your data is there.

You need that data to hand, he went on: 

You want that data to be there to power your CRM interactions, to personalize that marketing email, personalize that customer service interaction so that agent know who he or she is talking to. And you want the AI to be right there too, not off to the side, not [have to] switch over to some other AI interface. The AI needs to have that data. You want the AI to do some things for you, to make you more productive, to maybe write that email response to a customer or summarize the customer support issue. All of that working together, but powered by metadata. 

All this matters because the AI revolution is a data revolution, argued Patrick Stokes, Salesforce SVP of Product Management: 

You cannot have great AI without great data. We have islands of disconnected data. We have data all over the place. We have it in in our CRM, we might have it in our ERP, we've got it in all sorts of different places. But we've also got all sorts of different types of data. We've got something that we would call structured data, which is data that you would typically find in a database or a spreadsheet, but we've also got something called unstructured data. Unstructured data is things like PDFs, knowledge articles, chat scripts, transcripts, social media content, audio recordings - this is all unstructured data. We want to get access to that data. We'd like to have that data to provide context for our AI. 

Data Cloud now works with unstructured data. Now you can connect all of your unstructured data to Data Cloud and we use a vector database to effectively index all of that data and make it available for automation, for search and ultimately for Einstein copilot.

Data Cloud Vector Database and Einstein Copilot Search will both be in pilot in February next year, just as Einstein Copilot enters general availability. 

My take

An important addition to the Data Cloud story that clearly resonated with the Salesforce audience at Dreamforce in September.