Salesforce Deepens AI Integration, Targets Employee Productivity for Initial GTM Enhancements
Salesforce recently made some big announcements about its AI Cloud and how it was going to augment the capabilities of each of its other clouds, especially its Sales, Marketing, Service and Commerce clouds. But at its last month’s Dreamforce event in San Francisco, the company took the AI concept and ran with it — with AI being the central theme of this year’s conference.
We can’t really get into all of the AI announcements across all of the clouds, and we certainly will dig into the go-to-market focused additions in a subsequent brief. But for now, let’s dig into the core announcement: the new Einstein 1 platform and underlying components.
Einstein 1 Brings It All Together
In addition to the new platform, the company announced the pilot of Einstein Copilot, an out-of-the-box conversational AI assistant built into the UI of every Salesforce application. Think of Copilot as a ChatGPT-like prompt right in the UI, which allows users to ask questions in natural language and receive answers that are grounded in secure, proprietary company data from Data Cloud and the Salesforce application clouds. According to Salesforce, Einstein Copilot also proactively takes actions and offers additional options beyond the user’s query — for such use cases as providing a recommended action plan after a sales call, checking a consumer’s order status, or changing the shipping date.
Providing recommended actions is phase one, and a smart move. As generative AI is still a new concept, businesses might not want to automate the follow-up actions derived from AI just yet, nor provide AI with direct access to customers. Thus, tackling enablement is both adds value and clearly enhances productivity.
A point not to be missed is what Salesforce calls “Grounded Prompts.” By using Copilot for generative AI use cases, Salesforce can allow for both manual and dynamic improvement of results by adding more contact or company-specific data points into the prompt, for example. This allows for improved and more usable results. In addition, by using the Grounded Prompts inside the Salesforce platform, it can be done in a more secure fashion (see below on the Trust Layer).
The company is also piloting the new Copilot Studio— which itself leverages more generative AI capabilities to help users build custom use cases into their Salesforce instances. There are some interesting components, such as Flow — which takes MuleSoft’s data integration and workflow capabilities and makes them easier to build via natural language prompts instead of coding or a configuration kit.
It’s Only as Good as the Data…
Salesforce made two important announcements at Dreamforce aimed at making the use of AI more effective regarding the results and insights from conversational AI, but also for the continued development of a more broad and cost-effective data backend for Salesforce customers.
First, the company announced that Data Cloud is pre-integrated with the core business clouds — it is now free to use for any Enterprise user, and they can also utilize two free Tableau Creator licenses to consume the analytical insights generated by AI Cloud. There is a limit, however. Users can only create up to 10,000 unified profiles before having to pay Data Cloud fees. While this seems like a lot of records, when you think of Enterprise scales and the fact that B2B buying committees have at least five-to-seven individuals — hitting that 10,000 profile limit isn’t too difficult.
The second set of announcements had less fanfare, but may be quite important. Salesforce announced continued expansion of its Data Cloud partner ecosystem. As enterprises build out both their own large language models and AI strategies, Salesforce may have some critical data, but they certainly don’t have all of it. “Doing the work” around AI from a data integration perspective can be a costly and performance-limiting endeavor. Salesforce is doing the right thing by making it easier for users of Snowflake, Databricks, AWS and Google to hook up third party data lakes more easily with Salesforce. Salesforce can still be used as a customer data platform (CDP) and/or a data lake, but at scale it makes sense to be utilizing third party data providers, both due to costs and to ensure a broader set of data beyond what’s in the CRM.
It’s All About Trust
The new Einstein 1 announcements underscore a concept Salesforce has been developing for a long time: the Einstein Trust Layer. The most important aspect of the Trust Layer is that it shields sensitive data that is being used in generative AI prompts from being consumed by public or other large language models. Many large and regulated enterprises are dipping their toes into the AI waters but still need guidance on compliance and privacy — Salesforce’s Trust Layer could be a differentiator, as many other providers lack the platform capabilities to offer such a solid stance on providing a safe approach to using generative AI.
The injection of AI into everything is an inevitability, so Salesforce is smart to take a bullish position. We like how the company has made use of AI capabilities more seamless inside the core cloud products and has focused on ‘productivity’ use cases as low hanging fruit. We also really like how Slack can be a unifying UI for building an AI-powered, full journey engagement model for enterprises. Allowing all customer-facing employees to gain access to customer data usually locked inside Sales or Marketing Cloud instances via Slack can go a long way to building solid Converged Growth organizational procedures and policies.
Additionally, Salesforce is smart to be bandying about the “trust” concept all over its AI initiatives. Today, many uses of AI — especially generative AI — are going on unsupervised or without strategy inside enterprises around the world. For regulated industries, and for geographies with strict data privacy and stewardships mandates (with serious fines associated with each misstep), guardrails are needed. Salesforce is positioning itself as a trusted advisor both in terms of helping internally regulate how companies deploy generative AI in a safe and secure manner, but also performing a lot of the data masking and safety measures as a service.
Many other CRM and MarTech providers simply lack the breadth of functionality, services and a core platform to do what the Einstein Trust Layer can do for companies deploying an internal AI strategy. When you combine the ease of use that Slack can provide for all users in terms of interacting with AI prompts, to melding CRM and other Data Cloud information to improve AI performance with Grounded Prompts, to the data and infrastructure layers that MuleSoft and Tableau bring to the table to promote hyper-automation and workflow building with zero code — no one else really stacks up.
The devil, of course, is in the details of execution. While a lot of what we saw at Dreamforce was “live” and either already readily available or becoming available this fall — the full seamlessly embedded, “everything working together” vision Salesforce presents at its marquee customer event is still a bit more “art of the possible” than everyday use case. Salesforce customers are still going to have license access to lots of these components, which they may not have today and which will increase costs and complexity for larger enterprises. The key is how Salesforce continues to evolve not just the product portfolio, but its overall pricing models to make sure value is received on both sides as AI makes it easier, if not necessary, for more individuals in the company to access and make use of Salesforce data and features.