Salesforce Agentforce pricing in 2026: what conversations really cost.

Salesforce Agentforce is priced per conversation, around two dollars each, with Flex Credits and a Data Cloud dependency behind it, so usage forecasting decides the bill more than headcount ever does.

This guide is for procurement and Salesforce leaders sizing Agentforce in 2026. Pair it with the Agentforce licensing guide and the Salesforce Practice page so the usage model and the deal align.

How is Salesforce Agentforce priced in 2026?

Agentforce breaks the seat model. It bills per conversation, with Flex Credits as the prepaid alternative. The unit is usage, so the forecast is the cost case.

What counts as a conversation?

A conversation is one interaction session between a user or customer and an agent. Volume across service, sales, and internal use is what drives the bill, not the human seat count.

• Service agents: customer interactions can run at high volume.
• Sales agents: targeted use, lower conversation counts.
• Internal agents: employee facing use, variable volume.

How do Flex Credits work?

Flex Credits are a prepaid pool that agent actions draw down. They let you commit a budget and spend it flexibly, but they make a usage forecast essential rather than optional.

What drives the Agentforce bill?

Three factors set the real cost. Salesforce publishes the Agentforce model on its own Agentforce pricing pages, which is the reference for the per conversation rate and credit options.

Why does Data Cloud matter to the cost?

Most useful agents are grounded in your data through Data Cloud, which prices on its own consumption model. The true Agentforce cost includes that Data Cloud usage behind the agents.

How do you control the cost?

Control comes from forecasting and staging. Model conversation volume, size credits to a realistic curve, and track the Data Cloud usage that sits behind the agents.

• Forecast volume: estimate conversations before you commit.
• Stage rollout: add agents in waves, not all at once.
• Track Data Cloud: watch the consumption behind the agents.

What to do next

1. Estimate conversation volume across service, sales, and internal use.
2. Compare per conversation cost against the seat equivalent.
3. Size any Flex Credit pool to a realistic usage curve.
4. Include Data Cloud consumption in the full cost case.
5. Stage the agent rollout in waves to test real volume.
6. Design lean agent flows so actions stay light.
7. Negotiate the conversation rate and credit terms together.

Frequently asked questions

How is Salesforce Agentforce priced in 2026?

Agentforce is priced on a per conversation model, with a published rate around two dollars per conversation, and is also available through Flex Credits that draw down as agents act. The conversation count, not a per user seat, is what sets the bill, so usage forecasting matters more than headcount.

What is a conversation in Agentforce pricing?

A conversation is a single interaction session between a user or customer and an Agentforce agent. Because billing is per conversation, the volume of agent interactions across service, sales, and internal use is the figure that drives cost rather than the number of human seats.

What are Flex Credits in Agentforce?

Flex Credits are a prepaid consumption pool that agent actions draw down. They give buyers a way to commit a budget and spend it flexibly across agents, but they also make forecasting essential, since an underused pool wastes spend and an overused one triggers top ups.

Does Agentforce require Data Cloud?

In practice, most useful Agentforce deployments depend on Data Cloud to ground agents in your data. Data Cloud is priced separately on its own consumption model, so the real Agentforce cost includes the Data Cloud usage behind it.

How do you control Agentforce cost?

Forecast conversation volume before committing, size Flex Credits to a realistic usage curve, and track the Data Cloud consumption that sits behind the agents. Buyers who commit a large credit pool without a usage model tend to overspend.

Is per conversation pricing cheaper than per user?

It depends on usage. For high volume automated service, per conversation can cost more than seats, while for targeted internal use it can cost less. Model your expected conversation volume against the seat equivalent before committing.