Salesforce’s major AI pivot with Agentforce
Customer relationship management leader Salesforce is urging customers to give up on creating “DIY” copilot bots and instead leverage AI agents it is making available on its platform.
In a keynote address at the company’s annual Dreamforce conference in San Francisco yesterday, Salesforce CEO and founder, Marc Benioff, said Saleforce had been trialling Agentforce, a system for creating AI agents, which go beyond chatbots that dispense information, to automate complex tasks.
He later told journalists that he brought forward the launch of Agentsforce after seeing the dramatic improvements in productivity being achieved by customers, including Disney.
“The copilot world has been kind of a hit-and-miss world,” he said from the stage of the Moscone Centre in San Francisco.
“It's been a world where customers have some said, look, hey, I got these copilots, but they're not exactly performing as we want them to.
“We don't see how that copilot world is going to get us to the real vision of artificial intelligence, of augmentation of productivity, of better business results that we've been looking for. We just don't see copilot as that key step for our future.”
AI agents have been talking about for years and many companies are developing them. But Salesforce has an edge in that it hosts much of the data that underpin customer service, sales, marketing, and e-commerce operations. Salesforce has its own Einstein generative AI copilot to surface information held in Saleforce databases like Data Cloud.
An AI agent answers a customer service query. Source Salesforce
But Agentsforce creates a simple, low or no code way to instruct agents how to undertake tasks, such as processing a sales order via a chatbot, or in the case of restaurant reservation service OpenTable, handle booking enquiries and alter reservations.
Benioff claims that internal benchmarking shows that Agentsforce AI agents are more accurate and less prone to hallucination abd data leakage than competitors from the likes of Microsoft, Google, and OpenAI. That’s because the reference data is all held within Salesforce, so is the trusted source of truth powering customer interactions. External sources of data can also be shared with Salesfoerce via a “zero copy” system that doesn’t duplicate data, but draws it from its original source for the purpose of a transaction.
Clara Shih, CEO of Salesforce AI, said the Agentsforce platform offers over 100 industry-specific actions and allows for customisation using a tool called Agent Builder.
The Agent Builder used to programme AI agents. Source: Salesforce
She explained that there are five key elements of Agent Force: roles, data, actions, channels, and trust and security. The platform uses the Atlas reasoning engine to create plans, evaluate them, and refine them until it achieves the goal.
Agent Force leverages reinforcement learning from customer outcomes to continuously improve its performance.
Dreamforce featured demos from the likes of Saks 5th Avenue, OpenTable, and Bombadier to showcase how they are using AI agents. Benioff said he was using the conference to give visiting customers the opportunity to build their first AI agent. His goal is to have one billion people interacting with Salesforce agents by February 2025, the start of the company’s next financial year.
“We have flown in 4,000 of our own Salesforce engineers to work with you and help you get this done, and we've created these launchpads, and it's going to be awesome.”
AI Agents-related news from Dreamforce:
NVIDIA strategic partnership: A new strategic collaboration to advance AI Agent innovation and develop advanced AI capabilities for the enterprise and autonomous agent and avatar experiences. Both Salesforce and NVIDIA will work to deliver optimised predictive and generative AI workflows, accelerated data processing and retrieval augmented generation (RAG) and new types of AI agents.
Advanced AI capabilities tailored for enterprise use
Development of autonomous agents that enhance customer interactions
Creation of AI-driven avatar experiences for more engaging workflows
Agentforce Partner Network: The Agentforce Partner Network debuts today with industry leaders Amazon Web Services, Google, IBM, Workday, and more delivering specialised agents and intelligent agent actions that organisations can embed across their apps, customer channels and employee productivity tools.
Tailored AI agents autonomously act on unique business challenges.
Seamlessly integrates agents across platforms to improve efficiency.
Pre-built agents enable fast deployment and scalability.
Extended IBM partnership: Salesforce and IBM have extended their long-standing partnership to develop autonomous AI agents that organisations in regulated industries like financial services, manufacturing, telecommunications and insurance can deploy using their own proprietary data, controls and on-premises systems. This partnership allows businesses to leverage IBM’s Granite AI models within the Salesforce ecosystem, ensuring compliance with industry-specific regulations while accessing advanced AI capabilities.
Enables secure access to mainframe and Db2 data for AI tasks.
Automates tasks within compliant guardrails using Agentforce agents.
Integrates IBM’s Watson AI into Slack for advanced automation.
Google Agent ecosystem: Salesforce and Google Cloud today announced an expanded partnership to make work easier, with a new capability to build AI agents in Salesforce and Slack that utilise Gemini models via Google Cloud’s Vertex AI, and which securely leverage customer data from Salesforce and Google Cloud. Joint customers will be able to build Agentforce agents that work across Salesforce, Slack, and Google Workspace, as well as build custom AI Apps for Slack with Google Cloud Vertex AI.
Enterprise-grade security and privacy
Enables more seamless and efficient work across platforms
Peter Griffin attended Dreamforce as a guest of Salesforce
