Salesforce supercharges Agentforce with embedded AI, multimodal support, and industry-specific agents

Salesforce is updating Agentforce by embedding AI into every aspect of the platform in the hope that the no-code platform will move from pilot projects to production-grade systems that are ready for real-world digital execution.

The new capabilities added to Agentforce combine AI capabilities with Salesforce’s already deeply integrated ecosystem, spanning everything from customer data to workflows and analytics, said Dion Hinchcliffe, lead of the CIO practice at The Furutum Group.

The new capabilities include Agentforce for employee use cases, Agent Surfaces, Web Search in Agentforce Data Library, multimodal support, Instruction Adherence, and new features in Agentforce for Sales.

The first major AI upgrade was made to the Sales Cloud via Agentforce, which the company said will now reduce manual labor in managing sales pipelines by keeping CRM data accurate and updated by automatically suggesting or applying opportunity field updates, including “Next Steps” and “Stage” progression based on insights from customer conversations.

Additionally, it will help enterprises with sales development by supporting automated outreach across “Contacts and Person Accounts” in addition to “Leads.”

Further, sales administrators will be able to preview agent-generated messages in Agent Builder to ensure accuracy, plus multi-language support for global reach, Salesforce said.

The CRM software provider has also extended Agentforce and its functionality to employee use cases, where it will provide guidance, execute workflows, and surface knowledge to any employee by providing contextual and personalized experiences across Salesforce Lightning Experience, Salesforce Mobile, and Slack.

Increased scope of Agentforce

Amalgam Insights’ chief analyst Hyoun Park said the employee use cases capability is Salesforce’s efforts to increase the scope and reach of Agentforce and turn it into the “default experience” inside Salesforce.

He said Agent Surfaces is similarly designed to bolster Agentforce’s prowess by enabling admins and developers to create better agent experiences for users by making agents respond with visuals and media based on the context and the channel where they are deployed.

Designing better agent experiences, according to Salesforce, becomes possible as Agentforce now has multimodal support, which means users can provide audio, video, and images as additional context along with text to generate more accurate responses.

The Summer ’25 release of Agentforce also saw Salesforce make advancements to Sales Coaching by adding support for six new languages — French, German, Italian, Japanese, Spanish, and Portuguese, and on-demand availability to sales personnel.

Other updates to Agentforce include the ability for users to allow their agents to access sources on the public web, in turn expanding their knowledge base, and a capability named Instruction Adherence that acts as an additional monitoring and guardrails tool.

This tool, according to Salesforce, allows admins and developers to easily detect an agent’s adherence to its topic and instructions with AI-generated scoring.

Ahead of the curve

Analysts suggest that Salesforce’s steady stream of updates has helped Agentforce stay ahead of the curve.

“Where Azure’s Copilot Studio and Google Vertex AI Agents focus on developer-led customization, Salesforce is building deeply embedded, no-code agents that live inside core business functions,” Hinchcliffe said.

“Agentforce isn’t trying to be everything everywhere. Instead, it’s trying to be indispensable where the enterprise already works. And it’s a much broader audience,” Hinchcliffe added.

Industry versions and updates to Data Cloud

Salesforce has also launched new industry versions of Agentforce — Agentforce for Education to aid in solving student queries and booking activities, Agentforce for Financial Services for service assistance and customer service, Agentforce for Life Sciences to help with clinical site management, and Agentforce for Public Sector to help with complaint summarization and resolution.

The industry-specific capabilities that Agentforce now supports should provide users with greater guidance on how to incorporate agent usage into routine and time-consuming tasks, said Amalgam Insights’ Park.

The Summer ’25 release also saw the CRM software provider add new features to its Data Cloud offering in order to help enterprises manage their data more effectively.

Key among these updates are AI Tagging and Classification and RAG 2.0, which help in automatic labelling of data, metadata using AI-recommended tags and enhance agents’ accuracy by indexing extracted data with metadata and providing citations with responses, respectively.

Other updates to Data Cloud include unstructured connectivity enhancements and policy-based governance for better control of data access.

Enterprise teams, according to Salesforce, can use the unstructured connectivity enhancements feature to ground Agentforce agents in more unstructured data with intelligent document processing and new connectors, including Web Crawler, Site Map, and Zendesk.

“RAG 2.0, and other Data Cloud updates are underappreciated breakthroughs in my view: They let enterprises trust, audit, and scale AI agents without building custom governance frameworks,” Hinchcliffe said.

This week also saw Salesforce revising prices for Enterprise and Unlimited Editions, with Agentforce add-ons, effective August 1.