• Salesforce’s Agentforce lets businesses build custom AI agents.
• The platform leverages both proprietary and third-party AI models for maximum flexibility.

In the “CXO AI Playbook,” Business Insider delves into real-world examples of AI adoption across industries, business size, and technological frameworks. Featured companies share insights in the AI Playbook into the difficulties they hope to address using AI, the internal decision-makers driving initiatives, and their strategic aims for the technology’s future.

Salesforce is well-known for its customer relationship management (CRM) software, which is used by over 150,000 businesses, including Amazon and Walmart. It also owns Slack, a widely-adopted communication tool. The firm is now amplifying its AI capabilities with Agentforce, a platform that enables its clients to develop and deploy AI agents that automate tasks like generating sales data reports and summarising Slack conversations.

Identifying the problem: Evolving AI for dynamic needs

AI isn’t new territory for Salesforce. In 2016, it launched Einstein, an AI feature integrated in its CRM platform. Einstein was capable of handling scripted tasks, similar to Agentforce’s basic functions. However, advancements in generative AI opened new possibilities, like decision-making and better natural language comprehension.

The evolution began with Einstein GPT, which later became Einstein Copilot. By early 2024, it had transformed into Agentforce, offering prebuilt and customisable agents to meet a wide range of perceived demand.

“We recognised that our customers wanted to either extend the agents we provide or create their own custom AI agents,” said Tyler Carlson, Salesforce’s VP of Business Development.

Building the foundation: Key players and partnerships

Agentforce leverages Salesforce’s proprietary Atlas Reasoning Engine, developed by its engineering team. The platform also integrates with AI models from external partners such as OpenAI, Anthropic, Amazon, and Google, enabling businesses to harness a diverse set of technologies.

To showcase its capabilities, Salesforce uses Slack as a platform to highlight AI agents in action. Currently in beta, Agentforce’s Slack integration aims to bring automation directly to where users spend their time. “Slack is valuable for employee-facing agents because it makes their capabilities easily accessible,” Carlson explained.

AI in action: Smarter agents for complex tasks

The Atlas Reasoning Engine employs ReAct prompting which enables AI agents to break down problems step by step, evaluate outcomes, and refine their approach iteratively. This leads to more precise responses and personalised support, from answering customer inquiries to scheduling tasks autonomously.

The technology integrates Salesforce’s proprietary LLMs while API calls ensure compatibility with third-party models, so offering clients flexibility. “The beauty of agentic systems is that they’re not tied to a single LLM,” Carlson noted.

Salesforce has strict data privacy regulations when using external LLMs, such as limiting data retention and implementing protections to filter inappropriate responses.

Empowering clients: Custom AI agents at scale

Clients can use tools such as Agentbuilder to create and deploy tailored AI agents that act on predefined triggers. For example, an agent might sort emails and forward them to the relevant departments based on the content’s intent. Furthermore, retrieval-augmented generation enables agents access to internal company documents, making them capable of answering complex, organisation-specific questions.

One notable example is Salesforce’s work with Workday to develop an AI-powered employee service agent that can answer HR-related inquiries using a company’s private policies.

Measuring success: Adoption and outcomes

Salesforce gauges the success of Agentforce by tracking its clients’ achievements. Carlson cited early positive results, such as a 90% resolution rate for customer inquiries handled by AI agents.

The company’s long-term aims for Agentforce include widespread deployment in its CRM systems and Slack. Salesforce also aims to see AI agents handle more significant workloads and resolve complex interactions.

“By next year, we envision an expansive ecosystem of partners and thousands of agent skills and topics available to our clients,” Carlson said.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.