How Salesforce Agentforce AI Agents Work Across Sales, Service, Commerce in 2026 

Salesforce Agentforce is Salesforce’s autonomous AI agent platform. It lets agents reason, act, and complete tasks across sales, service, marketing, commerce, and IT, using a company’s own trusted data and governed by the Einstein Trust Layer. Since launching in August 2024, Agentforce has gone through several major releases. The most recent, Agentforce 360, became generally available in October 2025. The pricing model has also changed, moving from a flat $2-per-conversation fee to a flexible, action-based credit system. This guide covers what Agentforce actually does today, what it costs, where it fits, and how to implement it without the cost overruns that catch a lot of teams off guard.

What Is Salesforce Agentforce?

Agentforce is Salesforce’s platform for building and deploying autonomous AI agents. These aren’t chatbots that wait for a human to act on their suggestions. They take real actions inside your business systems: updating records, resolving cases, qualifying leads, processing orders, and more.

Salesforce now markets the underlying technology as the Agentforce 360 Platform, built across four layers:

• Agentforce 360 Platform. The reasoning and orchestration layer, including the Atlas Reasoning Engine, Agentforce Builder, and Agent Script.
• Data 360 (formerly Data Cloud). Unifies structured and unstructured business data so agents have accurate context to work from.
• Customer 360 Apps. Sales, Service, Marketing, and Commerce Cloud, where agents actually do their work.
• Slack. Increasingly the place where employees query data and trigger agents conversationally.

A rule-based bot follows a script. A copilot waits for a human to act on its suggestions. Agentforce agents can plan multi-step work on their own, call external systems through MuleSoft, and hand off to a person only when a task falls outside their guardrails.

What’s New in Agentforce 360

If your last read on Agentforce was from 2024, the product has moved on quite a bit. The biggest shift is Agentforce 360, announced at Dreamforce 2025 and now generally available. It added:

• Agentforce Builder. A conversational, “vibe-building” workspace for creating and testing agents without separate build, test, and deploy cycles.
• Agent Script. A human-readable scripting language that lets teams mix deterministic business rules with flexible AI reasoning, so agents behave predictably on the tasks that matter most.
• Agentforce Voice. A native voice layer that replaces IVR menus with real-time, low-latency conversations, integrated with contact center platforms like Amazon Connect, Five9, Genesys, and Vonage.
• Intelligent Context. Automatically extracts structure from unstructured documents (PDFs, emails, transcripts) so agents reason over more than just CRM fields.
• Multi-model support. The Atlas Reasoning Engine now works with OpenAI, Anthropic, and Google Gemini models, not just one proprietary model.
• AgentExchange and open standards. A partner marketplace and open protocols for connecting Agentforce to third-party agents and tools.

Salesforce has also rolled out purpose-built solutions like Agentforce Contact Center, which unifies voice, digital channels, and CRM data, and expanded Agentforce for Commerce with intent-aware product search. As of early 2026, Salesforce reports Agentforce has crossed $1B in annualized revenue, which says something about how quickly enterprises are adopting it.

How Salesforce Agentforce AI Agents Work

Every Agentforce agent runs on three things working together:

1. Trusted data. Pulled from CRM records and unified through Data 360, including the newly supported unstructured sources.
2. Reasoning. The Atlas Reasoning Engine plans the steps needed to complete a task, now configurable through Agent Script for more predictable behavior.
3. Governed action. The Einstein Trust Layer enforces data security, access controls, and auditability before any action executes.

In practice, this means agents don’t just generate a reply. They can resolve a case, update an opportunity, schedule a meeting, or finalize an order, with every action observable and reversible through Salesforce’s governance tools (Agent Health Monitoring, Observability, and Session Tracing).

Agentforce Use Cases Across Business Functions

Agentforce for Customer Service

Service agents resolve inquiries, troubleshoot issues, and escalate to humans with full context when needed. Agentforce Contact Center now unifies voice and digital channels so agents and people work from the same data, instead of relying on stitched-together point solutions.

Agentforce for Sales Teams

Sales agents qualify leads, prospect around the clock, and schedule meetings. They now operate inside a Sales Workspace that brings agents, analytics, and predictive insights into one hub, including Slack, where reps are increasingly managing pipeline without switching tools.

Agentforce for Marketing Operations

Marketing teams use Agentforce to draft and preview campaign briefs, identify audiences, build journeys, and optimize engagement. Much of this is now conversational and happens directly inside Slack.

Agentforce for Commerce and Retail

Beyond order processing and fraud prevention, Agentforce Commerce now includes Intent-Aware Search, which understands shopper intent in natural language instead of relying on static keyword rules. That addresses one of e-commerce’s biggest conversion blockers.

Agentforce for Operations, Finance, and IT

Agentforce IT Service meets employees in Slack, Microsoft Teams, or a portal to handle troubleshooting and access requests on its own, escalating only what genuinely needs a person.

Pre-Built Salesforce Agentforce Agents

Salesforce continues to ship ready-to-deploy agents that organizations customize rather than build from scratch:

• Service Agent. Resolves customer inquiries around the clock with seamless human escalation.
• Sales Development Representative (SDR) Agent. Engages prospects, answers questions, and schedules meetings across languages and channels.
• Sales Coach Agent. Runs reps through pitch practice and objection handling.
• Personal Shopper Agent. Drives product recommendations and conversion in digital commerce.
• Campaign Agent. Generates briefs, identifies audiences, and optimizes marketing performance.
• Agentforce IT Service and Contact Center. Newer additions that extend the same agentic model to internal operations and unified contact centers.

Who Should Use Salesforce Agentforce (and Who Shouldn’t)

Good fit:

• Enterprises already using Data Cloud or otherwise unified customer data
• High-volume sales, service, or commerce operations where action-based costs scale with clear ROI
• Organizations with defined AI governance and data readiness

Not an ideal fit yet:

• Small teams with limited data maturity or fragmented systems
• Organizations only looking for a basic scripted chatbot
• Businesses without a clearly defined automation use case, since consumption pricing punishes vague deployments

Agentforce vs. Copilots, Chatbots, and Rule-Based Automation

The distinction between these three is something Salesforce talks about a lot in its own positioning, and it’s also backed up by independent analysis. The short version: these tools differ primarily in who carries the responsibility for completing a task.

	Design intent	Acts without human input	Typical use
Chatbots	Follow pre-defined scripts	No	FAQ deflection, simple query routing
Copilots	Assist a human who stays in control	No	Drafting, summarising, surface-level suggestions
Agentforce	Execute autonomously within guardrails	Yes	End-to-end task resolution across systems

The key distinction is that copilots are reactive: a human initiates, and the tool assists. Agentforce agents are designed to perceive a trigger (a new lead, a support ticket, a contract expiring), plan the steps needed, and execute them across your systems without waiting for someone to act first.

That said, the lines are blurring. Microsoft Copilot Studio has moved in the direction of more autonomous agent behavior over the past year. So the table above reflects design philosophy more than a hard technical ceiling for any of these tools.

Salesforce Agentforce Pricing

Agentforce pricing has evolved significantly since launch and continues to change as Salesforce introduces new licensing tiers, credit models, and bundled editions. Rather than publish numbers that may already be out of date by the time you read this, we recommend going straight to the source for the most accurate and current information.

You can find the full breakdown of models, credit costs, and edition comparisons on the official Salesforce Agentforce Pricing page. If you’d like help modelling which option makes sense for your usage volumes and existing Salesforce licenses, get in touch with the Cloud Odyssey team and we can walk through it with you.

Salesforce Agentforce Implementation Considerations

Agentforce adoption isn’t plug-and-play. Successful rollouts tend to need:

• Data quality and Data 360 readiness. Agents are only as accurate as the data they reason over.
• Security, governance, and compliance alignment, including how the Einstein Trust Layer is configured.
• Integration with Sales Cloud, Service Cloud, and Commerce Cloud.
• Change management, since agents change how employees and customers interact with your business, not just what software runs in the background.
• Usage monitoring from day one, using Digital Wallet and Agent Health Monitoring to catch cost or performance issues early.

How Cloud Odyssey Helps Implement Salesforce Agentforce

Turning Agentforce on is the easy part. Getting measurable ROI out of it is not. Cloud Odyssey works with organizations to:

• Identify high-impact Agentforce use cases instead of deploying agents everywhere at once
• Design secure, scalable AI architectures aligned with Data 360 and governance requirements
• Model Flex Credits versus Conversations versus user-license costs against real usage forecasts
• Align Agentforce with multi-cloud Salesforce environments across Sales, Service, Marketing, and Commerce Cloud
• Connect Agentforce to external systems via MuleSoft Agent Fabric for secure, governed automation across the enterprise

You can see how this plays out in our CurrencyFair customer query handling case study, or read about Agentforce for small businesses if you’re trying to figure out fit for a smaller org.

Frequently Asked Questions

Is Salesforce Agentforce still priced at $2 per conversation?

Not exclusively. The $2-per-conversation model is still available, but it’s now one of several options. Most new deployments use Flex Credits, priced at $500 per 100,000 credits, where each standard agent action costs about $0.10.

What’s the difference between Agentforce, Agentforce 2dx, Agentforce 3, and Agentforce 360?

Agentforce launched in 2024 as the original autonomous agent platform. Agentforce 2dx (2025) added proactive, multi-step agents. Agentforce 3 (mid-2025) focused on interoperability and governance at scale. Agentforce 360 (late 2025 onward) is the current generation, adding Agentforce Builder, Agent Script, Agentforce Voice, and Intelligent Context.

Can Agentforce use AI models other than Salesforce’s own?

Yes. The Atlas Reasoning Engine now supports OpenAI, Anthropic, and Google Gemini models, so organizations can choose which model reasons and plans for their agents.

Is there a free way to try Agentforce?

Yes. Salesforce Foundations gives Enterprise Edition and above customers 200,000 Flex Credits and 250,000 Data 360 credits at no cost. It’s a permanent tier, not a time-limited trial.

Does Agentforce replace human agents entirely?

No. Agentforce is built to handle high-volume, repetitive work and escalate complex or sensitive cases to humans, who end up spending more time on higher-value, high-empathy work rather than being replaced outright.

Conclusion

Salesforce Agentforce has moved well past its 2024 launch framing. With Agentforce 360, multi-model reasoning, native voice, and a pricing model that’s been rebuilt from the ground up, it’s a different and more capable product than the one Salesforce introduced two years ago. The strategic case for autonomous, governed AI agents inside Salesforce hasn’t changed. What has changed is the level of control you get, the breadth of use cases on offer, and how carefully you now need to model costs before committing. Get the data foundation, governance, and pricing model right, and Agentforce becomes a genuine force multiplier rather than another line item to manage.