Salesforce Launches eVerse: A New AI Agent Simulator

Salesforce has entered a new phase of AI advancement with the introduction of eVerse, a new simulation environment built to train voice and text agents. Instead of analyzing real customer calls or conducting manual testing, eVerse allows companies to train their agents in a virtual world that is realistically chaotic, and entirely unpredictable.

This is a big deal since today’s businesses are completely dependent on AI agents. These bots are fast and powerful, but they often get confused by small changes in how people speak, type, or behave. With eVerse, Salesforce hopes to close that divide by providing AI with a place where it can learn, make mistakes, and get better over time.

At its heart, eVerse is a training ground. Think of it as a giant simulation world where AI agents can talk to fake customers, deal with tricky situations, and learn how to respond better the next time.

Salesforce built eVerse because today’s AI often struggles with “jagged intelligence.” This means an AI can be extremely strong in some areas but suddenly fail at a simple question that a human would answer in seconds. For example, riddles, sarcasm, humor, or broken speech still confuse many modern AI models.

Salesforce eVerse targets these blind spots and helps companies build agents that are more natural, more context-aware, and more reliable.

Most businesses want AI agents that talk like humans, solve problems fast, and don’t break down in real-life situations. But training an AI to handle thousands of different accents, noise levels, emotions, and conversation styles is extremely hard.

This is where eVerse shines. Instead of waiting for customers to call with issues of every kind, companies can simulate those moments and see how their agents handle them. Then they can iterate, improve and test again without the cost and time waste.

eVerse uses a three-step cycle: Synthesize, Measure, and Train. This cycle repeats until the agent reaches the quality a business needs.

Salesforce shared that eVerse played a major role in building Agentforce Voice, its new voice AI. Even today, many customers prefer calling, so the system had to handle real-world challenges like noise, accents, interruptions, and emotional conversations.

With eVerse’s realistic simulations, Salesforce ran thousands of test calls to train and refine Agentforce Voice. This early testing helped teams fix issues quickly and ensure enterprise-level stability.

Madhav Thattai, COO of Salesforce AI, explained,

Salesforce is still working out the pricing model for eVerse. As the pilot continues, the team is evaluating how many tokens each test run should use and how that consumption will translate into billing.

Madhav Thattai offered clarity on how things function today:

Customers can access agentic features through their existing Agentforce subscriptions, or they can choose a usage-based option where they’re charged for the actions their agent performs in real operations. The core platform tools are included; only the live actions executed by the agent are billed.

At this stage, Salesforce eVerse is available only to a select group of customers as part of an early pilot rollout while Salesforce finalizes the commercial details.

Salesforce’s eVerse is a game changer for agent AI development in the recent years. It provides an intelligent, controlled, and realistic environment to develop agents that better understand human behaviors.

Leveraging synthetic data, stress testing, and reinforcement learning, eVerse enables organizations to develop agents that are prepared to confidently emerge from real-world challenges.

While there are some pricing and long-term question marks, eVerse seems like an AI built for the future. More businesses are coming online through the pilot program: we’ll get a sense of how well it scales and just how much it defines the next generation of AI agents.