Salesforce is significantly expanding its AI capabilities with Agentforce, a comprehensive platform designed to help its 150,000+ clients—including Amazon and Walmart—build and deploy AI agents that automate complex business tasks. This evolution represents a major leap from the company’s earlier Einstein AI service launched in 2016, which offered more scripted automation capabilities.
The transformation from Einstein to Agentforce reflects Salesforce’s response to the rapid advancement of generative AI technology. What began as Einstein GPT, then evolved into Einstein Copilot in early 2024, has now become Agentforce—a platform offering both prebuilt and fully customizable AI agents. According to Tyler Carlson, Salesforce’s Vice President of Business Development, the company recognized that customers wanted flexibility to either extend existing agents or build entirely custom solutions.
At the heart of Agentforce lies the Atlas Reasoning Engine, proprietary technology developed by Salesforce’s engineering team that uses ReAct prompting to help AI agents reason through complex problems. This approach enables agents to break down challenges into components, provide step-by-step solutions, and evaluate results after each action—looping through different reasoning paths until reaching satisfactory outcomes. The concept mirrors OpenAI’s o1 model and its chain-of-thought prompting technique.
Agentforce’s capabilities are impressive and diverse: agents can generate written reports from sales data, create conversation summaries in Slack, handle customer inquiries autonomously, schedule appointments without human intervention, and answer employee questions about HR policies. The platform integrates with third-party AI models from OpenAI, Anthropic, Amazon, and Google, though Salesforce has implemented strict data retention policies and safeguards to protect client information.
The Agentbuilder tool allows clients to create custom agents that respond to specific triggers—like receiving emails or completing sales—without direct employee prompts. Using retrieval-augmented generation (RAG), these agents can answer questions based on internal documents and private company data. Salesforce and Workday have already collaborated on an AI employee service agent for HR inquiries.
Early results show promise, with Carlson reporting that some implementations have achieved 90% resolution rates for customer inquiries. Salesforce’s integration with Slack (currently in beta) is particularly strategic, surfacing AI capabilities where employees already spend their time. Looking ahead, Carlson envisions Agentforce developing into a robust ecosystem with “thousands upon thousands of agent skills” available to customers by next year.
Key Quotes
We had to recognize that our customers want to either extend the agents that we ship or they want to build their own customer agents
Tyler Carlson, Salesforce’s Vice President of Business Development, explained the company’s strategic shift toward offering customizable AI agents rather than one-size-fits-all solutions. This recognition drove the evolution from Einstein to the more flexible Agentforce platform.
The reason Slack is interesting for these employee-facing agents is that it surfaces these automations and capabilities to users where they’re spending their time
Carlson highlighted the strategic importance of integrating AI agents directly into Slack, emphasizing that effective AI adoption requires meeting users in their existing workflows rather than forcing them to adopt new platforms.
This is what I find most interesting about agentic systems. They’re not monolithic. They’re not a single LLM.
Carlson described Agentforce’s multi-model architecture, which allows access to AI models from OpenAI, Anthropic, Amazon, and Google alongside Salesforce’s own technology. This approach represents a significant departure from single-vendor AI solutions.
By next year I want to see a larger and more open ecosystem of partners. I want Agentforce to have thousands upon thousands of agent skills, topics, and a lot more that our customers can take advantage of.
Carlson outlined Salesforce’s ambitious vision for Agentforce, positioning it not just as a product but as a comprehensive platform ecosystem that could become the standard for enterprise AI agent deployment.
Our Take
Salesforce’s Agentforce launch is strategically brilliant, positioning the company at the forefront of the next wave of enterprise AI—autonomous agents rather than simple chatbots or copilots. The evolution from Einstein to Agentforce demonstrates how quickly the AI landscape is changing; what seemed cutting-edge in 2016 now appears primitive compared to reasoning-capable agents.
The multi-model approach is particularly noteworthy. Rather than betting everything on proprietary AI, Salesforce is building an orchestration layer that can leverage the best models for specific tasks. This hedges against any single AI provider’s limitations while giving customers flexibility—a smart strategy in a rapidly evolving market.
However, the 90% resolution rate claim deserves scrutiny. While impressive, the complexity and context of those resolved inquiries matters significantly. Simple queries are easier to automate than nuanced customer problems requiring judgment and empathy. The real test will be whether Agentforce can handle edge cases and complex scenarios without frustrating customers or requiring frequent human intervention. If Salesforce achieves its vision of thousands of agent skills, it could fundamentally reshape how enterprises approach workflow automation.
Why This Matters
Agentforce represents a pivotal moment in enterprise AI adoption, signaling the shift from simple automation to truly autonomous AI agents capable of complex reasoning and decision-making. For businesses, this technology promises significant productivity gains—the reported 90% customer inquiry resolution rate suggests AI agents can dramatically reduce workload on human employees while maintaining service quality.
The broader implications extend beyond Salesforce’s ecosystem. By creating a platform that integrates multiple third-party AI models rather than relying on a single proprietary system, Salesforce is establishing a blueprint for enterprise AI architecture. This multi-model approach acknowledges that different tasks may require different AI capabilities, representing a more mature understanding of AI deployment.
For the workforce, Agentforce exemplifies the ongoing transformation where AI handles routine tasks while humans focus on higher-value work. The integration with Slack—where millions of workers already collaborate—makes AI agents accessible and practical rather than theoretical. As more companies deploy similar agentic systems, the nature of knowledge work will fundamentally change, requiring workers to develop new skills in managing and collaborating with AI agents rather than performing routine tasks themselves.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources: