Four reasons why the new Data Cloud is so strategic for Salesforce

This year's Dreamforce has seen the launch of a significantly upgraded Data Cloud as part of a rebranded Salesforce platform, now known as the Einstein 1 platform. How big a move is this for the CRM giant? According to David Schmaier, the company's President and Chief Product Officer, "We did a pivot earlier this year, saying we want all-in on AI and data." Here's what we've been able to glean this week about what makes this new platform so strategic.

1. Data is separating from application stacks

The wider picture is that all enterprises are currently looking at their data and wondering how to extract it from all the application-centric silos that have grown up around it over the years. The rapid growth of a new generation of data aggregation vendors including Snowflake, Databricks, Confluent and others is testament to this trend. For Salesforce, the challenge is to expose and unify data across both its own applications and from other sources, so that it can keep pace with this wider trend.

This is the background to what Patrick Stokes, EVP of Product and Industries Marketing, says is the biggest evolution in the platform “in the last decade or so.” What makes it so significant is that the re-architected metadata layer has overcome the limitations that have always been inherent in running on top of an Oracle transactional database architecture. This has always created difficulties in connecting other types of data that run better on other architectures. Most notably for Salesforce, this includes the data that runs in its own Commerce Cloud and Marketing Cloud, as well as other unstructured data sources. Stokes explains:

When you're working within the Salesforce platform, and you're constrained to Oracle as your underlying storage mechanism, you're going to run into things like, it's not really great for unstructured data, it's not really great for trillions of records, it's not great for objects that have thousands of fields. So that's what we've been working on as we totally reconstructed that core metadata layer to further abstract the underlying storage mechanism, to go from something that complements our transactional database underneath which is Oracle, with a more of a data lakehouse architecture, which is the data lake. And then we connect the data lake to that metadata framework to bring it in ...

We've effectively tricked the Salesforce platform into now, reading from a data lake in the same way it does Oracle, and it can't tell the difference.

Whereas the previous iteration of Data Cloud was effectively a Customer Data Platform (CDP) that focused on marketing, commerce and customer data, the new version has a broader scope and connects that data into all of the development and admin tooling of the Salesforce platform. He explains:

Your data sitting in the data lake benefits from all of the capability of the Salesforce platform just like it does in the transactional side in Oracle. That means that Data Cloud is now not just a really good CDP, it's now a data lake which will be used in sales and service use cases. But it also means that we can start to fundamentally move some of our higher-scale consumer products like Marketing and Commerce onto the platform … All of this consolidation coming to the platform is what makes the platform new and different. Connecting all of this data via that new data mechanism of Data Cloud is just awesome.

2. AI needs a governance layer

There’s been enormous excitement all this year about the potential of generative AI, tempered by concern about whether businesses can trust this exciting new technology. The advantage of unifying all data into a single Data Cloud platform is that it then becomes possible to add a single governance layer to mediate interactions between AI and data. In the new platform architecture, this is known as the Einstein Trust Layer.

One of the key elements in this mediation layer is known as ‘grounded prompts’, which are essentially questions posed to a generative AI model in order to obtain an answer. In the examples shown during Dreamforce by Clara Shih, CEO of Salesforce AI, the Einstein 1 platform allows these prompts to be grounded in data objects found in the Data Cloud. Some of them are transactional, such as customer or product name, while others are unstructured data such as policy documents. Grounding prompts in this real-world data gives the AI model more guidance so that it is more likely to produce an accurate answer. This kind of AI governance is crucial for building trust in the technology. Schmaier references Geoffrey Moore, author of Crossing the Chasm, the classic business book about technology adoption, who has been advising Salesforce:

I think what's really going on here is really a second chasm, where we're in a generative AI early adopter cycle, where everybody's kicking the tires and doing their pilots. I think what a lot of people are finding is, if you just take whatever software you have, and you plug in OpenAI, it doesn't really get you there. And I think that's not good enough.

3. AI reinforces the business case for modernization

A vendor’s latest offerings are always several steps ahead of where many of its customers are at. There are Salesforce customers that still haven’t adopted its Lightning interface, which was first introduced a decade ago. Many customers have multiple instances and face challenges connecting across them. A large number haven’t yet migrated to the AWS-based Hyperforce infrastructure. It’s not so much that they don’t want to, it’s just that they don’t have a compelling reason for incurring the cost and effort of upgrading.

At a time when many CEOs are eager to see their IT leaders formulate an AI roadmap, AI is becoming a forcing factor for IT modernization. With Data Cloud easing data consolidation, the new platform helps build a business case for modernization. Schmaier comments:

Data Cloud, I couldn't be more in love with this product ... This is truly an incredible unlock. Companies have tens, or hundreds or thousands of systems. You could have done this in the old days, and some of my prior companies did — MDM systems and ESBs, plug it all together — but it was still a complicated task.

Whereas now ... you just plug it in and the data just streams and flows together and coordinates according to the canonical data model, so that you can have the account data so you actually know where your customers are. That's an app that's pretty important for any company. So now you can do that right out of the box with Data Cloud.

4. Customer data stays in Salesforce

At a time when data is separating from applications, as I mentioned above, the danger for Salesforce — and indeed for any other vendor — is that customers will decide to keep their data in a completely separate data store, and then find it much easier to switch at will to an alternative application. Or it could be an opportunity for Salesforce to make a land grab beyond its traditional CRM territory and become the platform of choice for back-office and operational data too. I put this to Schmaier and his response was that Salesforce is partnering with data companies rather than aiming to replace them:

We're not trying to be the one lake to rule them all. We could. By the way, Data Cloud is built on the same open source technologies that Snowflake and Iceberg and Spark and others are, so it actually could be. But that's not our goal. Our goal right now is to really focus on solving this for Salesforce customers.

I do believe that people will bring a lot of operational data into Data Cloud, or join it through zero-copy integration. Because what you want to see is the customer with the payment history, with the order history, with the shipping history, aggregated by customer. So that's again, why all the data companies want to work with us, because we have all this valuable data.

At the same time, he doesn’t see customers moving their CRM data to a different platform:

We think we live in a multi-lake world. Our goal is, we're trying to make it easy for people to do this, I would go beyond CRM, I'd say all across the front- and mid-office is our ambition ...

People can choose to be a computer hobbyist and take one of those data warehousing technologies, build their own star schema, map our objects, and then export the data through those APIs and populate the data, and then keep that in sync with our objects as we change them in our three-release roadmap per year.

After a while they'll [think], 'Hey, why don't we take things easy and have this all come out of the box for our people?’ I would predict, in two or three years, that's going to be the way most people are going to do it. Because you don't want to have this stale lake house that you custom-built. You'll want the one that we have, that takes all the operational data from the front- and mid-office. And it also powers again, the AI. That's why the data and the AI piece is so important.

My take

There's a lot riding on the success of Data Cloud.