At Dreamforce, Salesforce unveiled Agentforce Sales, a new vision for customer relationship management built around autonomous agents designed to help sales teams move faster and work more efficiently.
The product marks an early attempt to rethink how selling happens inside Salesforce, and it is already reshaping internal workflows. Yet modern sales work rarely unfolds in a single application.
Sales professionals often juggle research and drafting in ChatGPT, record updates in Salesforce, emails in one tab, calendars in another. This constant switching has a cost. Inside the industry, it is sometimes called the “Toggle Tax,” the mental friction that accumulates when attention is fragmented across tools.
Salesforce is now trying to reduce that burden by bringing its CRM directly into ChatGPT.
Agentforce Sales app connects OpenAI’s models with live Salesforce data. The integration allows sales staff to view account details, pipeline information and customer histories without leaving the ChatGPT interface. Using natural language, users can query records, update fields and trigger sales workflows.
Agentforce Sales was initially introduced as part of Salesforce’s push into autonomous sales agents that can handle routine tasks. The ChatGPT app extends that strategy beyond Salesforce’s own platform into a third party environment that many sales teams already rely on for research, planning and writing.
Context inside ChatGPT
The Agentforce Sales app links ongoing ChatGPT conversations directly to a company’s Salesforce instance. Account details, lead statuses and opportunity records can be pulled into prompts as contextual information, subject to the same user permissions that apply inside Salesforce.
Sales users can ask for lists of new or uncontacted leads and receive live, filterable results drawn directly from the CRM. By eliminating the need to copy and paste information between systems, the app reduces both preparation time and the risk of outdated data shaping decisions.
The app also supports conversational updates to CRM records. A user can instruct the system to change the status of an opportunity, assign a lead to an engagement flow or log notes from a recent call. Those changes are written back to Salesforce without requiring the user to open the CRM interface.
Combining data sources
Salesforce is positioning the ChatGPT integration as a way to blend internal CRM data with external signals that OpenAI’s models can access. The app takes into account lead scores, pipeline health and broader market or news indicators when ranking which prospects a seller should prioritise.
The goal is to present sales representatives with a more complete picture of customer potential, one that reflects both internal activity and changing external conditions. The system can generate draft account plans outlining priorities, key performance indicators, risks and growth opportunities, then store those plans inside Salesforce.
Routine outreach tasks can also be delegated through natural language commands. Lower priority leads can be routed into automated engagement flows, allowing human sellers to focus on larger or more complex deals.
Customer reaction
Asymbl, a Salesforce customer, has begun using the app across its sales organisation.
“Using ChatGPT without Salesforce was frustrating because the conversation stopped short,” said Brandon Metcalf, CEO and Founder, Asymbl. “We could talk about opportunities, but we couldn’t move them forward. With the Agentforce Sales app in ChatGPT, our reps can prioritise deals, plan accounts, and update Salesforce directly from the conversation. It will change how our sales team works and boost productivity.”
At its core, the app is designed to reduce the need for sales teams to move between multiple browser tabs and applications throughout a sales cycle. It brings CRM workflows into the same interface where many users already draft emails, research prospects and prepare meeting notes.
Security and governance
Salesforce says that all data exchanged between ChatGPT and Salesforce is routed through what it calls the Agentforce Trust Layer, a controlled channel that governs how information flows between the two systems.
According to the company, customer data processed through the app remains subject to existing Salesforce security standards. ChatGPT can access only the data a user is already permitted to see in Salesforce, and that data is used solely as context for responses.
The Trust Layer also evaluates AI generated recommendations before they reach users. Salesforce says the intent is to ensure that suggestions align with company policies and that proprietary business information stays within approved boundaries.
Meeting sellers where they work
Salesforce has framed the launch as part of a broader effort to embed Agentforce across the tools sales teams already use, including Salesforce itself, Slack and now ChatGPT.
The Agentforce Sales app for ChatGPT is now available in open beta. It can be installed from the ChatGPT app directory by customers using the Agentforce for Sales Add-on or Agentforce 1 Edition.
Kris Billmaier, GM and EVP of Agentforce Sales at Salesforce, said the company expects sellers to handle more of their daily pipeline work inside ChatGPT as a result of the integration. “With the Agentforce Sales app in ChatGPT, we’re bringing Salesforce into the tools sellers already use,” said Billmaier. “Now conversations with the Agentforce Sales app in ChatGPT understand your customers, move deals forward, and get real sales work done.”