9621 Agnes Crossing, Lake Suzanneview, New Mexico Island 84604-9295.
At Dreamforce, Marc Benioff shares his vision for the future of AI for businesses. Copilots are a thing of the past, autonomous agents are the future. And he’s
At Dreamforce, Marc Benioff shares his vision for the future of AI for businesses. Copilots are a thing of the past, autonomous agents are the future. And he’s making it happen, because in his opinion, no one is better suited to do it than Salesforce.
At Dreamforce, Salesforce CEO Marc Benioff presents a vision. According to him, the future of AI will consist of agents: autonomous digital AI bots that can correctly perform a variety of tasks.
“We have already helped you with your rollout Sales Representative“, says Benioff, “and then you Marketing powerYou Analytical powerYou Data power and now you agentforce.”
Agentforce is new, but then again it isn’t. Saleforce is retroactively making previously introduced agents such as the Einstein Service Agent part of Agentforce, which was introduced as a concept on the eve of Dreamforce. Convenient, because it means Benioff can announce Agentforce version 2.0 in one fell swoop, which is based on the new Atlas Reasoning Engine.
Agentforce is, as the name suggests, a collection of AI-powered agents. These agents are small applets that are integrated with your Salesforce environment and in particular the data in the Data Cloud, but can also communicate with third-party integrated environments such as AWS, Google and IBM.
The core of Agentforce 2.0 and the agents is the Atlas Reasoning EngineSalesforce developed it itself and is designed to mimic how people evaluate and answer a question. First, a query is evaluated and possibly refined with further questions, then Atlas retrieves the relevant data, after which the engine creates an execution plan. Atlas uses the Get Augmented Generation (RAPPER). “By the way, RAG was invented here at Salesforce,” Benioff boasts.
Agents are not co-pilots, Benioff is clear about that. “Copilot is the next Clippy,” he thinks. “Customers don’t see the value. They want to skip the co-pilot step.” According to Benioff and Salesforce, copilots have missed their opportunity. It seems they aren’t delivering the productivity gains promised. Agents have to solve that. What about the Einstein copilot, then? It’s being upgraded and becoming Agentforce.
Copilot is the next Clippy.
Marc Benioff, CEO Salesforce
The agents differ from today’s AI applications in several areas. Most notable is their autonomous functionality. Just like an intern, for example, they can successfully complete certain tasks without human intervention. Consider lead triage and follow-up or customer support.
On stage, we see a demo that illustrates the functionality. Someone calls the customer service line of clothing store Saks, where an AI agent named Sophie answers. Sophie can speak (in English) and has access to various Saks systems, such as customer data, product data, and inventory data.
When someone calls and asks for a garment that is too small, Sophie knows immediately which garment it is. The agent looks at previous orders from the customer, who is immediately identified by the phone number. Sophie then offers to send the item in a new size or suggests the customer pick it up from a nearby location.
The entire conversation takes place in natural language and Sophie immediately understands what is important. In contrast to a classic chatbot, the conversation does not follow a predetermined structure. Unlike a copilot, the AI does not support a human support agent: Sophie does everything herself. The Atlas engine makes this possible.
Agents live in the Salesforce platform in addition to the Customer 360 applications. An agent consists of five components:
Security is ensured by the Einstein Trust Layer. This system monitors the active ingredients so that they work within clearly defined limits and do not hallucinate. For example, if a Sophie could not find a clear solution to the problem that the garment is too small, but did find an answer, as LLMs sometimes do, the trust layer will prevent this.
Benioff himself is already very impressed with the way the functions work together within the Salesforce platform. “Agentforce is the biggest breakthrough we’ve ever had in technology, and I think it’s the biggest breakthrough I’ve ever seen in AI,” he says with a certain sense of superlatives.
Salesforce has been working on agents and the Atlas engine for some time, but Benioff only decided to make Agentforce the topic of the conference a few weeks ago. “The keynote was completely ready, as usual,” he tells the assembled press. “And it worked well. But in my discussions with customers, I found that they have both high expectations and a lot of frustration with AI.” This realization must have come sometime before the announcement of the quarterly results in late August, because at that time he was already remarkably negative about Copilot.
Benioff then decided to develop a new keynote, which was only completed on the eve of Dreamforce. “We want to break the hypnosis,” he says. “The AI implementations with basic LLMs are do-it-yourself science projects. Companies don’t need do-it-yourself.”
“Enterprise AI is much more than an LLM,” added Clara Shih, CEO of Salesforce AI. “It was a big confusion for 18 months. An LLM needs to be surrounded by the entire technology stack in an organization to enable productivity.”
With Agentforce, Salesforce claims to offer this stack. “Only Salesforce offers what you need,” Benioff continues. “We have an unfair advantage with the Salesforce platform. Customers already have their data in the Data Cloud, they’ve already built flows and integrated MuleSoft, the whole platform is already there.”
Benioff is not afraid to use big words. Although Salesforce is not really inventing anything new at Dreamforce, Salesforce is proposing a useful combination of technologies that can be easily implemented in a platform that many companies already have. Salesforce itself is introducing various agents with relevant functionality as part of its industry-specific cloud offering and is particularly encouraging customers to develop agents themselves.
We hear that agents can help companies increase capacity during peak times without the need for additional staff. Agents take over boring, mundane tasks, freeing people up to do more inspiring work. We even see that agents prevent burnout because they give employees extra oxygen.
Will everything go smoothly? Benioff expects it will. He backs up his promise with concrete goals: “We want to have thousands of customers with Agentforce by February 1st. Next year, at the next Dreamforce conference, we’re aiming for one billion consumers who have interacted with an agent.
We want to have thousands of customers using Agentforce by February 1st.
Marc Benioff, CEO Salesforce
Agentforce has potential, especially for companies that are already deeply embedded in the ecosystem. Salesforce itself is now removing legacy code from its platform as quickly as possible. Everything is being moved to Salesforce Core, which makes applications speak the same language. This is required for legacy applications and acquisitions such as Slack and Tableau. This is the only way the AI agents can integrate seamlessly into the entire platform.
An organization with a lot of data in Data Cloud and Salesforce Foundations can get started quickly. A conversation with an agent costs $2 per hour (or less). In many cases, this is easy to justify. After all, what is more efficient: having a paid employee send a personalized follow-up email to a hundred leads or automatically assigning this task to an agent?
Still, Benioff is quite brutal in his communication. Copilot has been retired, so to speak, and all other AI implementations are being relegated to do-it-yourself projects that do not deliver value. In his opinion, Agentforce is the solution. In practice, generative AI has not been around for long. Everyone wants an immediate return on their investment, and Agentforce has the potential to quickly deliver results for customers who are sufficiently involved in the ecosystem.
On the other hand, it can be said that generative AI is very new and evolving very quickly. Nobody knows what the best implementation is, not even Benioff. “We’re taking a bit of a gamble here,” he admits. AI models are evolving, frameworks are maturing, and AI is increasingly being incorporated into an enterprise framework. For example, within Cortex AI, Snowflake also provides an AI implementation in a secure data cloud, equipped with advanced generation for retrieval.
Benioff and Salesforce tell the story more specifically, backing it up with concrete examples of early adopters. Agentforce is unique in this regard. Salesforce also states that thanks to the Einstein Trust Layer, the AI technology is reliable enough to work as an independent colleague rather than as a human assistant.
Source: IT Daily
As an experienced journalist and author, Mary has been reporting on the latest news and trends for over 5 years. With a passion for uncovering the stories behind the headlines, Mary has earned a reputation as a trusted voice in the world of journalism. Her writing style is insightful, engaging and thought-provoking, as she takes a deep dive into the most pressing issues of our time.
