Dreamforce 2026 Highlights: Every Major Announcement from the Keynote

If you’re searching for Dreamforce 2026 highlights or a full list of what Salesforce actually announced this year, here’s the short version: the event centered on one new product, AIforce, and everything else, Claudeforce, Slackforce, the new Agentforce agents, Koa, plugs into it.

Here’s the problem Salesforce built AIforce to solve. A sales rep asks ChatGPT what’s happening with a deal, and it has no idea, because it’s never seen the deal. IT stands up a chatbot for support tickets, and it can answer general questions but not “what’s the status of ticket 4021.” Marketing runs something on Claude, but it can’t touch the actual campaign data sitting in Salesforce. Every tool is smart in general and useless on specifics, because none of them are actually connected to the company’s real data.

Salesforce’s fix is a new interface layer that connects any AI model to a company’s actual Salesforce data, permissions, and workflows, wherever an employee already works, whether that’s Slack, Claude, or Salesforce itself.

Here’s every major announcement from the keynote, what it actually does, and what it means if you’re running on Salesforce.

Dreamforce 2026 Keynote Theme and Opening Moments

Benioff opened Dreamforce 2026 by naming a problem most enterprises quietly have. Enterprises today use multiple AI tools, but none of them talk to the company’s actual Salesforce data. Generative AI can sound impressive and still get the numbers wrong, because it isn’t connected to a source of truth.

Salesforce sees this as a major shift in how people will interact with business software. Instead of keeping AI inside one application, the same business data, workflows, and logic can work across different interfaces, including Slack, Claude, and Salesforce. The idea is that employees should be able to use AI where they already work without losing access to the business context behind their work.

This also connects to a bigger question Salesforce kept coming back to throughout Dreamforce, which is what all these new announcements and products actually mean for business results. AI creates ROI by increasing revenue or reducing costs. The products Salesforce announced are designed to make that possible by giving agents access to the data, systems, and workflows they need to turn AI capabilities into real business work.

Top Dreamforce 2026 highlights 

1. AIforce brings Salesforce’s data, workflows, business logic, permissions, security, and governance into whichever AI interface you’re already using.

2. Claudeforce puts Salesforce context and actions directly inside Claude, backed by 37 prebuilt sales skills.

3. Slackforce turns Slack into a working interface for Salesforce, creating records, coding with agents, and building live dashboards, all from inside a channel.

4. Agentforce expanded with job-ready agents and a long-horizon runtime, letting agents pursue goals over days or weeks instead of handling one task at a time.

5. Koa and the Enterprise AI Harness mark Salesforce’s push toward CRM-specific reasoning and a more controlled, governed foundation for enterprise AI.

AIforce: The Four-layer Architecture 

AIforce was the central product announcement of the keynote. AIforce sits at the interface layer. It makes that data, application intelligence, and agent capability available through dynamic experiences instead of a fixed Salesforce screen.

But AIforce isn’t working alone. Salesforce has connected it to four key parts of the platform, so it can access the right data, understand the business, and take action. These parts include:

• Data 360 brings together the company’s actual data and helps keep it clean and connected. That gives AI access to reliable information instead of pulling answers from disconnected sources.
• Customer 360 adds the applications, business processes, and logic that sit around that data. This is what helps AI understand what the data actually means in a business context.
• Agentforce brings in the AI agents that can act on that information. They are built for specific jobs, such as pipeline generation, customer service, or other business tasks.
• Then comes AIforce, which brings all of this together at the interface level. Instead of making employees switch between different systems or AI tools, they can interact with the data, applications, and agents through the interface they are already using.

But AIforce is not limited to Salesforce. It brings the same business context and AI capabilities into the tools people already use: Claude, Slack, and Agentforce.

Dreamforce 2026 Product Announcements

1. Claudeforce: Salesforce inside Claude

ClaudeForce puts Salesforce data and capabilities inside Claude, so users can get work done without constantly switching between tools. 

• It uses MCP to connect Claude with Salesforce.
• Claude can access only the Salesforce data the user already has permission to see.
• It comes with 37 prebuilt sales skills to get teams started.
• Setup takes just a few clicks, without separate technical configuration.

2. Slackforce: Slack as “the front door to Salesforce”

Slackforce turns Slack from a messaging app into a place where you can create records, code with agents, and generate live dashboards, all without opening Salesforce. 

At Dreamforce, Slackforce is announced with 3 core capabilities, including:

With Slackforce, enterprises can bring Salesforce data and business context into Slack and take action without switching between the two platforms. Slackbot can work within the same permissions and business rules already defined in Salesforce.

3. Agentforce: A new generation of job-ready AI agents

Salesforce announced a major expansion of the Agentforce agent portfolio at Dreamforce 2026, shifting the conversation from what agents can technically do to how much real business work they’re already doing.

These are the job-specific new Agentforce agents announced at Dreamforce:

These agents are only one part of the Agentforce expansion. Salesforce also introduced new capabilities to help enterprises control how agents behave, coordinate them across workflows, and improve their performance over time. These capabilities include:

• Agent Script is an open-source language that combines AI reasoning with rules, giving enterprises more control over how agents behave. 
• Multi-Agent Orchestration, now generally available, allows multiple agents to work together on the same business process, while Agent Optimizer helps identify underperforming agents and recommends improvements.

4. Agentforce Coworker: The interface inside Salesforce Lightning

Agentforce Coworker brings AI directly into the Salesforce interface, so employees can ask questions, find information, and take action without switching to another AI tool. Agentforce Coworker reached 100,000 users in its first 35 days, showing how quickly AI can become part of the Salesforce workflow.

At Dreamforce, Salesforce announced Coworker with capabilities including:

• Natural-language search: Ask questions about Salesforce data instead of manually searching through records.
• Agent handoff: Coworker can bring in specialized Agentforce agents when a task needs more than a simple answer.
• Custom AI Skills: Teams can teach Coworker how to handle specific business tasks and processes.
• Existing permissions: Coworker works within the user’s Salesforce permissions and business rules.
• Mobile access: The experience is available across desktop and mobile.

5. Builder Central: Building your own apps and agents without coding

Builder Central is Salesforce’s new AI-powered workspace for building custom apps and agents without writing code. It can generate data models, agents, flows, validations, and tests from a single request. It also connects the right agents to different parts of a workflow, so each task is handled by the agent best suited for it.

At Dreamforce, Salesforce demonstrated Builder Central with capabilities including:

• Natural-language building: Describe the outcome you need instead of starting with code.
• Existing Salesforce data: Reuses data and objects already available in the org.
• Automated development: Generates data models, flows, validations, tests, and other components.
• Agent coordination: Connects multiple agents so they can handle different parts of the same workflow.
• Custom business processes: Builds apps and agents around specific business requirements that prebuilt agents do not cover.

6. Koa: Salesforce’s First CRM Reasoning Model

Koa moves beyond using a general-purpose LLM as the reasoning engine behind Agentforce. It is built to reason through CRM tasks, not just generate answers. 

Salesforce trained KOA on synthetic scenarios based on real CRM workflows, including how agents should make decisions, use tools, and complete tasks step by step.

7. Trust, Governance, and the Enterprise AI Harness 

As agents get more freedom to make decisions, access data, and take actions on their own, enterprises need to keep track of what hundreds of agents are doing across the business. 

At Dreamforce, Salesforce announced the Enterprise AI Harness, a set of capabilities designed to keep agents secure, connected, and under control as they move into live business workflows.

Salesforce highlighted several pieces of this layer:

• Salesforce Guardian: Helps protect data across agent interactions, with zero data retention so customer data isn’t used to train the underlying models.
• Agentic Identity: Gives every agent its own identity instead of having it act as a person. Each action can be traced back to the agent and the person responsible for it, with permissions that can be limited down to specific fields and actions.
• Observability Center: Gives teams one view of what agents are doing across sessions, MCP calls, Apex, Flow, and errors, rather than checking different logs separately.
• Headless Experience Layer (HXL): Makes Salesforce data, business logic, APIs, and permissions available to AI interfaces without requiring users to work through the Salesforce UI.
• AI Control Plane: Gives enterprises a central way to discover, register, monitor, and govern agents across Salesforce and third-party platforms.
• Agent Fabric: It brings agents built on different platforms into one connected system, so enterprises can manage Salesforce and third party agents together instead of treating them as separate silos. With Agent Fabric, enterprise teams have visibility into what agents are doing, what they cost, where their data comes from, and whether they are following business policies.

So, building an agent is only half the job. Enterprises also need to know what that agent can see, what it can do, and whether they can trace its actions when something goes wrong. 

Other Dreamforce 2026 Announcements

Beyond the core AIforce announcements, Salesforce also expanded the same idea of bringing business context to AI across government, cloud platforms, and data. It announced:

1. Missionforce for Government and Regulated Industries 

Salesforce expanded Missionforce with new capabilities for secure government use, including a partnership that brings OpenAI models into Salesforce Government Cloud. Its new Policy Engine can turn approved policies into structured rules that agents can use when making decisions.

2. Deeper AWS and Google Cloud partnerships

Salesforce is making its data and agents available across both ecosystems. Amazon’s AI tools can access Salesforce context, while Agentforce can use models through Amazon Bedrock. Salesforce also expanded its Google Cloud integration, including Gemini Enterprise and plans to bring Commerce Cloud checkout into Google Search and the Gemini app.

3. Data 360 goes headless

Data 360 is becoming a common context layer for AI across the enterprise. Its Agent Context Engine gathers the right data for each request in real time, so agents can work with relevant business information without being given access to the entire dataset.

What Dreamforce 2026 means for Salesforce users

Dreamforce announcements are built around a bigger shift where AI agents can work with real business data, follow existing processes, take action across systems, and handle more complex work with less human intervention.

1. Start with Ready to Use Agents

Job ready agents such as Piper, Hunter, Fin, and Marshall already come designed for specific business functions. Teams can start with these agents, connect them to the processes they already run, and customise them around their needs.

2. Build trust as AI agents Gain Autonomy

As agents get more freedom to access data and take action, capabilities such as Salesforce Guardian, Agentic Identity, and Observability Center become important parts of the rollout. Enterprises need to know what an agent can access, what it did, and who is accountable for its actions.

3. Let Employees use AI where they Already work

AI does not have to live inside Salesforce. Through AIforce, Salesforce is taking the same business context into the interfaces employees already use, including Slack, Claude, and Lightning. The goal is to let people work with AI without giving up the data, permissions, and business context behind their work.

4. Measure Agent ROI

Enterprises can track agent performance, identify where an agent is struggling, optimise it, and decide whether it is delivering enough value to keep expanding its role.

5. Manage Agents as they Scale

As more agents enter the business, companies need a way to manage, monitor, and govern them. Agent Fabric and the Enterprise AI Harness provide the infrastructure to do that.

Wrapping up

Dreamforce 2026 points to a shift from running AI agents in silos to building a more connected enterprise, where agents can work with the same business data, context, workflows, and systems. 

So instead of each agent solving an isolated task, they can work together toward larger business goals, whether that means moving a deal forward, resolving a customer issue, improving a supply chain process, or helping a team achieve its KPIs. For enterprises, the focus now is on connecting these capabilities in a way that lets AI move beyond individual tasks and contribute to meaningful business outcomes.

Making this connected agent ecosystem work requires the right Salesforce architecture, integrations, and governance, and Cyntexa’s certified Salesforce and Agentforce consultants can help you put it into practice.

Book a meeting to move from siloed AI agents to a connected Agentforce ecosystem

AUTHOR

Vishwajeet Srivastava

Salesforce Data Cloud, AI Products, ServiceNow, Product Engineering

Co-founder and CTO at Cyntexa also known as “VJ”. With 10+ years of experience and 22+ Salesforce certifications, he’s a seasoned expert in Salesforce Data Cloud & AI Products, Product Engineering, AWS, Google Cloud Platform, ServiceNow, and Managed Services. Known for blending strategic thinking with hands-on expertise, VJ is passionate about building scalable solutions that drive innovation, operational efficiency, and enterprise-wide transformation.