Salesforce Agentforce

What is Salesforce Agentforce?

Salesforce Agentforce is a set of AI agents that run on Salesforce data and workflows for service and sales. Where a standalone chatbot answers from documents alone, an Agentforce agent reads the CRM record in front of it, grounds its reply in your data, and acts through Salesforce Flows. A service agent can look up a case and update it, and a sales agent can qualify a lead and log the next step, each inside the platform your team runs on.

Salesforce makes the product. The company launched Agentforce as part of its platform push into agentic AI, building on its Einstein and Data Cloud investments. Each agent draws on CRM data and the Trust Layer, so the same governance that covers your Salesforce org extends to the agent.

The audience is Salesforce customers. Organizations that keep service cases, sales pipeline, and customer records in Salesforce get the most from an agent that reads and writes that data. Teams outside the Salesforce ecosystem gain less, since the value comes from the CRM grounding and the native Flows.

Key features

Agentforce centers on features that tie an agent to your Salesforce data and let it complete work:

• ▸CRM-native agents: agents read and write Salesforce records, so a service or sales conversation uses the case, contact, and account data your teams already keep.
• ▸Data Cloud grounding: agents pull context from across your connected sources through Data Cloud, so answers reflect current customer data rather than a static help center.
• ▸Flows and actions: agents run Salesforce Flows to take steps such as updating a case, creating a record, or triggering a process, so they finish tasks in the platform.
• ▸Einstein Trust Layer: guardrails, data masking, and auditing wrap each agent, so sensitive data stays protected and responses stay inside policy.
• ▸Service and sales use cases: prebuilt agent patterns cover common service deflection and sales support work, so teams start from a template instead of a blank canvas.
• ▸Platform integration: agents live alongside your existing Salesforce clouds, so they share the same security model, users, and data.

The CRM grounding is what sets Agentforce apart. Because an agent reads live records and acts through Flows, it answers with your data and completes tasks in your systems. That capability depends on your Salesforce setup: an agent grounds well when your data is clean and your Flows are built, so preparation shapes the result.

How well does it work?

Agentforce performs well for organizations that keep their service and sales data in Salesforce. When an agent grounds in Data Cloud and acts through Flows, it can resolve common cases and support deals with the same records your team uses, which cuts manual work and keeps answers consistent with your data. The Trust Layer gives security and compliance teams the guardrails they expect from the platform.

The limits track the model. Agentforce is built for Salesforce customers, so its value falls for teams that keep data elsewhere. An agent grounds only in the data and Flows you connect, so gaps in your CRM become gaps in what the agent can do. Standing up clean grounding and reliable Flows takes admin and data effort, and complex or edge cases still route to a person. Teams that invest in data quality and clear guardrails see the strongest results.

A practical way to raise what an Agentforce agent resolves is to ground it in your highest-traffic data first, then build Flows for the common actions those cases need. Coverage of your top service and sales tasks drives the return.

Salesforce Agentforce pricing

Agentforce pricing starts from $2 per conversation. Rather than charge per seat, Salesforce ties cost to each agent conversation, so spend tracks usage. The final rate depends on volume, your Salesforce edition, and the add-ons you enable such as Data Cloud, which you confirm with Salesforce.

Because the rate varies with your contract, the tiers below describe how pricing is structured rather than fixed totals:

The per conversation model favors teams that want cost to track usage, though high volume can make the total hard to predict, so estimate your conversation counts before you commit. Because the rate ties to your wider Salesforce agreement, budget for a conversation with your account team and ask how a conversation is counted, what Data Cloud grounding adds, and how the rate scales, so you can compare the total against seat-based tools.

Who should use Salesforce Agentforce?

Agentforce fits Salesforce customers that want AI agents grounded in their CRM data and workflows. It suits these groups in particular:

• ▸Service teams on Salesforce Service Cloud that want agents to deflect and resolve common cases using live case and account data.
• ▸Sales teams on Sales Cloud that want agents to qualify leads, answer questions, and log next steps inside the pipeline.
• ▸Organizations with data across many sources that Data Cloud can unify and ground an agent in.
• ▸Enterprises with security and compliance needs that require the guardrails and auditing of the Einstein Trust Layer.

Agentforce is a weaker match for teams that keep their data outside Salesforce, or for small teams with light volume, where a lighter standalone tool would cover the job without the platform investment.

Alternatives and how it compares

Agentforce competes with a field of AI support and agent platforms. The right comparison depends on your stack and how much CRM grounding you need.

• ▸Zendesk AI: an AI layer built into the Zendesk support suite that resolves conversations and assists agents, a fit for teams on Zendesk rather than Salesforce.
• ▸Intercom Fin: an AI support agent that resolves conversations across chat, email, and social and prices per resolution, tied to the Intercom platform.
• ▸Sierra: a platform for conversational agents that take actions across voice and chat, priced on outcomes and independent of any single CRM.

Agentforce's edge is its CRM-native design: an agent grounds in your Salesforce data through Data Cloud, acts through Flows, and inherits the Trust Layer governance. If your service and sales data lives in Salesforce, Agentforce is built for that stack. If your data and support sit on another platform, a tool native to that platform may fit with less integration work, so weigh your stack alongside the per conversation cost.

Limitations and getting started

Be honest about the trade-offs before you commit. Agentforce's value rides on Salesforce, so teams that keep data elsewhere gain less, and an agent grounds only in the data and Flows you connect. The per conversation pricing can grow hard to predict at high volume, and standing up clean grounding and reliable Flows takes admin and data effort.

Getting started follows a clear path:

1. 1Identify the top service and sales tasks you want an agent to own, and note which Salesforce data and Flows each one needs.
2. 2Prepare your data, including Data Cloud connections, so the agent grounds in current, clean records.
3. 3Build or confirm the Flows for the actions the priority tasks need, so the agent can complete them.
4. 4Configure the Einstein Trust Layer with guardrails, data masking, and the cases you want a human to own.
5. 5Launch on a subset of conversations, review agent responses and handoffs each week, then widen coverage as the agent earns trust.

A staged rollout keeps risk low: start with one high-volume task and a small share of conversations, confirm the agent grounds and acts as intended, then add tasks as your numbers hold up. Because pricing tracks conversations, the early weeks tie spend to the work the agent does.