Salesforce: AI is the new UI

“AI is the new UI,” said Salesforce President Sarah Franklin at the Salesforce Connections marketing and commerce event in Chicago this week.

What this means, she said, is that the familiar experience of “pecking through” websites to find and learn about products will be replaced by generative AI-powered engagement that will take each prospect or customer on a journey based around their needs and interests — using their interactions, of course, as “prompts.”

Franklin believes this vision will become a reality very soon. “We are at the precipice of an AI revolution,” she said. “It is imperative that everybody pay attention right now.” The depth of change was underlined by her insistence that we stop thinking about applications and start thinking about models.

Throughout several keynotes, Salesforce demonstrated how users of its suite will be able to generate refined audience segments, create individual customer journeys and perform many other tasks by using natural language prompts within the course of normal workflows.

Generative AI across the suite

In support of this “AI is UI” concept, the main topic of discussion at Connections was the extension of Einstein GPT not only across all the clouds that make up the Salesforce suite — Marketing, Sales, Service, etc — but also to Slack and Tableau.

Marketing GPT will launch with segment creation using natural language prompts. Later will come email content creation, rapid identity resolution, dynamic product descriptions and the ability to leverage third-party solution Typeface to create visual content within the cloud. Note that the release of the individual capabilities will be staggered, mostly planned for Q3 2023 or early 2024.

Commerce GPT is aimed at delivering instant, personalized experiences at each stage of the customer journey, including not just text conversations but images and video. This supports Franklin’s vision of replacing online catalog shopping with an end-to-end interactive experience.

User responses to all the above will be pumped back into Data Cloud and used to support the continual updating of user profiles.

There will be an additional cost for Salesforce customers adopting these generative AI features. Details are not yet available, but should become clearer, Franklin said, at the Salesforce AI Day event to be hosted by Marc Benioff on June 12.

Capabilities for marketers

We asked Jay Wilder, VP of product marketing for Marketing Cloud, to dive deeper into some of the coming capabilities.

“One area within marketing that marketers like myself would really lean into would be around segmentation — being able to work with data more easily,” he said. “You have billions of rows of data about your customers and one of the big complaints I still hear is that it’s a very complex, manual process to get into that data, explore it, understand what’s there, and then ultimately generate segments that help drive a more personalized experience. We’re taking it to the next step where you can just literally describe in your own language what you’d like to see in that segment.”

In other words, Einstein GPT will turn a natural language prompt into a database query. “This really reduces the barrier of entry to using more sophisticated capabilities. Being able to segment faster doesn’t need me to understand how the data works; I just need to know how I think of it in my natural language,” he said.

“Then email for content creation,” he continued. “Subject lines, body copy — that’s piloting in October and coming in the spring. That will allow you to generate many different versions of your content, either to get to the right one that you want, or to do massive personalization at scale.”

In alignment with Franklin’s “visionary” comments, what Wilder sees coming down the road is a situation where marketers can begin with natural language descriptions of their strategic aims and targets, leaving it to AI to generate instantaneously the customer journeys to achieve the desired outcomes.

“Moving from more singular campaigns to more expressive journeys that are more responsive. It’s a win for companies, but I also think it’s a win for consumers because personalization at scale, for most, is not a reality today. This I think will help solve that.”

Are you getting the most from your stack? Take our brief 2023 MarTech Replacement Survey

Putting guardrails in place

But Salesforce isn’t quite leaving it all to AI. “It’s not just about doing all that,” he said, “it’s doing it with trust. A lot of our customers are very excited about this capability, they see it as assistive to what they’re doing today — but how do they make sure their data doesn’t leak out into the large language model? How do I make sure the results of the prompt are actually personalized to my customer rather than generic?”

The first step to making the process secure is to ensure that any data that passes into the LLM is almost instantly erased. “It goes ‘poof,'” said Wilder.

The second step: “Making sure there’s a marketer in the loop for everything so that it’s on-brand and on-voice.” Salesforce, he said, is alert to “challenges around hallucinations (AI inventing things) and confidentiality and embedding it in the workflow so I can use it in situ while doing my job on a daily basis.”

But isn’t that just the problem? Generative AI may be able to create everything from emails to ads to complex customer journeys, but doesn’t someone have to review and approve everything?

“Someone does,” he agreed, “and that’s really important.” The time savings, he insisted, remains real and significant. “It shifts the workflows from being about how the content gets created manually to how we review and tweak a body of work. There still is a very purposeful human-in-the-loop approach here, but the ability to quickly go from concept to ideation to research to fleshing out a program in literally a matter of moments is a far different experience from weeks of calls and meetings and cross-functional creative processes. The human-in-the-loop will still be there to make sure the content is approved, but I think it’s a good trade-off in terms of work.”

Expanding the partnership with Google Cloud

Another announcement at Connections was an expansion of the partnership between Salesforce and Google Cloud and specifically with Google BigQuery, the scalable enterprise data warehouse. Salesforce Data Cloud, said Wilder, is part of an open ecosystem. “We previously made announcements with Snowflake and with Amazon SageMaker as well as our tier-one advertising partners.”

There are two elements to the Google announcement, Wilder explained.The first is about Big Query. Food giant General Mills was a guest at the Connections marketing keynote: “General Mills, all of their data across a hundred brands and a global business lives in Google BigQuery. Traditionally you would have to copy and move that data [to use it in Salesforce]. Wouldn’t it be much better if we could see and use that data and just leave it in place?” That’s what zero copy data access allows and that’s what the partnership will support.

Secondly, the partnership will give zero copy access to Google’s Vertex AI to train AI models on Salesforce data. “In the old world, you would have to copy all that data into Vertex, run and train the models, and then copy it all back into Salesforce. No longer. With zero copy you can actually train the models on Salesforce data, leaving everything where it is.”

Like Franklin, Wilder sought to underline the imminence of these initiatives. “This is shipping quicker than I think most folks anticipate. These capabilities are coming to real production usage in the next few months.”