Salesforce bids to become copilot building platform of choice

The potential for generative AI copilots to transform businesses has created a lucrative opportunity for vendors vying to become the go-to platform for building AI assistants.

On Wednesday, Salesforce unleashed its latest offering in the race for the prize with Einstein 1 Studio, a new set of low-code/no-code AI tools for accelerating the development of gen AI applications. Central to Salesforce’s bid for enterprise gen AI development mindshare are Copilot Builder for creating custom AI actions for business tasks, Prompt Builder for building and activating custom prompts, and Model Builder for building or importing large language models (LLMs) and other AI models.

“There is a race under way to become the preeminent platform for building business copilots, and Salesforce is committed to leading the charge,” says Paul Nashawaty, practice lead and lead principal analyst at The Futurum Group, specializing in application modernization.

Nashawaty’s colleague Keith Kirkpatrick concurs, noting that Salesforce has all the tools to become the platform for building out and deploying gen AI assistants.

“They have a ton of data, due to their massive customer base, from which to train models, as well as the metadata, which is the data that describes or carries the attributes of that data, like flows, mappings, and business rules,” says Kirkpatrick, research director at Futurum.

David Menninger, SVP and research director at ISG’s Ventana Research, agrees that the race to become the platform for building business copilots is on, calling it “a mad dash.” But he’s not ready to call a favorite.

“Nearly every vendor of every application or technology platform is introducing copilot-style functionality,” he says. “I don’t think we’ll end up with just one.”

As Menninger sees it, a best-of-breed copilot platform will need to support a variety of LLMs, as well as be able to fine-tune them and incorporate small language model (SLM) capabilities to ensure its copilots adapt to an organization’s specific needs. It will also need to be capable of incorporating data from a variety of data sources, both internal and external to the enterprise.

“Finally, it will need strong governance capabilities to ensure the accuracy of responses, prevent biases, and protect privacy and intellectual property rights,” he says.

Salesforce makes its pitch

Salesforce Einstein 1 Studio builds on Salesforce’s previous work rearchitecting Data Cloud and the Einstein AI framework to create Einstein 1, an open platform for unifying Salesforce data to enable the development of gen AI-based applications. Einstein 1 Studio and the Data Cloud Spring ’24 release, also announced on Wednesday, seek to further enhance customers’ ability to unlock the power of their data with gen AI.

“Through initiatives such as the Einstein 1 Studio and Data Cloud Spring ’24, they are focusing on developers to deliver AI solutions as an enabler of rapid application development, as well as interacting with business lines to expand the technology,” Nashwaty says.

In this way, Einstein 1 Studio is a step forward in Salesforce’s efforts to democratize AI development, he says, pointing to features designed to streamline the creation and deployment of AI-driven applications, enabling developers to leverage advanced machine learning capabilities without “extensive expertise” in data science or AI algorithms. Meanwhile, Data Cloud Spring ’24 aligns with the company’s commitment to providing developers with access to actionable data sources, Nashwaty says.

“By integrating with various data sources, developers can leverage the vast troves of information available within Salesforce Cloud to fuel their AI initiatives and drive meaningful business outcomes,” he says.

Robert Parker, SVP of industry, software, and services research at IDC, sees the Einstein Studio 1 announcement as “a logical next step,” noting Salesforce’s “astute investments in their platform, from real-time customer data (Genie), AI (Einstein), and an augmented interface (copilot).”

But with adoption and results key to winning out in the enterprise, Parker points out an X factor that could help tip the race in Salesforce’s favor: “What was most interesting to me about this announcement was the launch of Trailhead courses, which will allow them to mobilize a large and committed developer ecosystem,” he says.

A crowded field

While Salesforce’s ultimate goal may be to offer a general-purpose business platform, Menninger says it’s still all about Salesforce data.

“Salesforce has always been an applications company first,” Menninger says. “They have introduced or acquired technologies such as MuleSoft, Slack, and Data Cloud, which could help them become more of a platform play, but it’s a long road. Einstein 1 Studio could help further their attempts to become a platform, but it is still initially targeted at Salesforce applications.”

Like Salesforce, other large software vendors such as SAP and ServiceNow are pushing their own gen AI copilots for business (Joule and Now Assist, respectively). All three face similar challenges when it comes to handling data that is external to their applications, Menninger says.

“Enterprises that are committed to one of these platforms would consider bringing in outside data, but organizations that are committed to multiple application vendors are less likely to do so,” he says. “Nonetheless, it is important for these vendors to be able to incorporate these other data sources; otherwise, their value becomes diminished since so many analyses and operations require data from across the enterprise.”

To that end, SAP yesterday announced enhancements to its Datasphere “business data fabric” to facilitate governance for non-SAP data assets within customers’ environments.

End game

In the end, there may not be one single winner in the race to become the copilot building platform of choice. Instead IDC’s Parker sees three enterprise strategies emerging in the business copilot platform space.

The first would involve siloed development, in which each business function uses a combination of models and tools to create its own portfolio of copilots. The second would be to align each business function’s copilot strategy around “a set of ‘super platforms,’ like Adobe for marketing, Salesforce Data Cloud for sales, SAP for finance, etc.,” he says.

The third strategy would be to adopt a single strategic partner (or a minimal number of partners), through which all copilot projects would be conducted.

“This takes longer, but allows you to assert more control,” he says, noting that IBM, AWS, Google, and Microsoft are the most likely partners for this approach.

Whatever the outcome, IT leaders no doubt have a lot vested in keeping a close eye on how this race plays out.