sforce: AI is now the new interface //

Sarah Franklin, Salesforce’s president, said that AI is the new user interface at the Salesforce Connections Marketing and Commerce event in Chicago last week.

She explained that this will replace the traditional experience of “pecking” through websites to learn about and find products with a generative AI-powered interaction that will guide each customer or prospect on a journey that is based on their interests and needs — using, of course, their interactions as “prompts.”

Franklin is confident that this vision will soon become reality. She said, “We’re on the brink of a revolution in AI.” It is important that everyone pay attention now.

Salesforce showed in several keynotes how its users will be able create refined audience segments, individual customer journeys, and perform other tasks using natural language prompts as part of their normal workflows.

Generative AI in the Suite

The main topic at Connections, in support of the “AI is UI”, concept, was the extension Einstein GTP to Slack and Tableau, as well as all the Salesforce clouds — Marketing, Sales and Service, for example.

Marketing GTP launches with segment creation based on natural language prompts. Later, email content creation will be followed by rapid identity resolution, product descriptions, and the ability for Typeface, a third-party solution, to create visual content in the cloud. The release of each capability will be staggered. Most are planned for early 2024 or Q3 2023.

Commerce GTP aims to deliver instant, personalized experiences for each stage of the journey. This includes not only text conversations, but also images and videos. Franklin’s vision to replace online catalog shopping by an interactive end-to-end experience is supported.

Data Cloud will use the user responses to the above questions to update the profiles of users.

Salesforce customers will have to pay an extra cost for these AI-generated features. Franklin stated that details are not available yet, but will be made clear at the Salesforce AI Day hosted by Marc Benioff, on June 12.

Marketers’ capabilities

Jay Wilder is the VP of Product Marketing for Marketing Cloud. We asked him to explain some of the upcoming capabilities.

He said that segmentation is an area of marketing where marketers such as himself would like to focus more. “Being able work with data easier” would be a priority. “You have billions and millions of rows of customer data and one of the biggest complaints I hear is how difficult it is to explore, understand, and ultimately create segments that drive a more customized experience. “We’re going to take it to the next level where you can literally describe in your language what you would like to see in this segment.”

Einstein GTP can also convert a natural-language prompt into a query. This really lowers the entry barrier to more advanced capabilities. “I don’t have to understand the way data works to be able to segment quicker; I only need to know what I mean by it when I speak my native language,” said he.

He continued, “Then you can use email to create content.” Subject lines, body text — this will be piloted in October and available in spring. This will enable you to create many different versions, either to find the content you are looking for, or to personalize it at scale.

Wilder, in agreement with Franklin’s comments about “visionary” marketing, sees a future where marketers will be able to describe their strategic goals and targets using natural language, then leave it up to AI to create the customer journeys that achieve the desired results.

“Moving away from singular campaigns and towards more expressive journeys which are more responsive. This is a great win for the companies, but it’s also a good win for the consumers, because for many, personalization on a large scale isn’t a reality. This will, I believe, help to solve that.”

Installing guardrails

Salesforce doesn’t leave everything to AI. He said that it’s more than just doing everything. It’s about doing it with confidence. Many of our customers find this feature very useful, and they think it will help them in their work. But how can they be sure that their data won’t get into the large-language model? How can I ensure that the prompt is personalized for my customer and not generic?”

To make the process more secure, the first step is to ensure all data entering the LLM are erased almost immediately. Wilder said, “It’s a poof.”

He said that Salesforce is aware of “challenges surrounding hallucinations and confidentiality, and has embedded it into the workflow to allow me to use it on a regular basis while I am doing my work.”

Isn’t this the real problem? But isn’t that the problem?

He agreed that “someone” does, and the time savings are real. It shifts workflows away from how content is created manually and towards how we review, tweak, and improve a body work. The human in the loop approach is still very much present, but it is now possible to go from concept through ideation and research to fleshing out an entire program within a few moments. This is far more efficient than weeks of meetings and calls with cross-functional teams. “The human-in the-loop is still there to ensure that content is approved. But I think it’s worth the trade-off.”

Google Cloud expands its partnership

A second announcement made at Connections involved the expansion of the Salesforce-Google Cloud partnership, specifically the Google BigQuery enterprise data warehouse. Wilder said that Salesforce Data Cloud is a part of an ecosystem. “We made previous announcements with Snowflake, Amazon SageMaker and our tier one advertising partners.”

Wilder explains that the Google announcement has two components. The first one is Big Query. General Mills, the food giant, was a guest speaker at Connections Marketing keynote: “All of General Mills’ data from a hundred brands as well as a global business is in Google BigQuery.” In the past, you had to copy and transfer that data in order to use it with Salesforce. It would be better to just use the data in its current location and not have to copy it. That is what zero-copy data access allows.

Second, the partnership gives Google’s Vertex AI zero-copy access to Salesforce data to train AI models. In the past, it was necessary to copy the data into Vertex and run the models. Then, copy the data back into Salesforce. Now you don’t have to. “With zero copy, you can train models on Salesforce data and leave everything else where it is.”

Wilder, like Franklin, wanted to emphasize the imminent nature of these initiatives. This is shipping faster than most people expect. In the next few months, these capabilities will be used in real production.

The original post Salesforce AI: the new UI was published on MarTech.