Agentforce Licensing in 2026. Conversations, actions, and Data Cloud.

Salesforce Agentforce is the company's conversational AI agent layer for Service Cloud, Sales Cloud, and Marketing Cloud. It prices on a conversation metric, with a separate per action ledger that meters every tool call the agent makes against the underlying Salesforce data.

The conversation list price sits at 2 USD per conversation in 2026. The agent action ledger consumes Einstein Requests at 0.10 USD per request. Most production deployments require a Data Cloud subscription as a prerequisite, with a per Customer 360 Profile metric that adds 5 to 15 percent to the run rate.

Read this alongside the Salesforce knowledge hub, the Salesforce services page, the Salesforce licensing guide, the utilization calculator, and the Vendor Shield subscription.

How Agentforce pricing works

Agentforce sits on top of the existing Salesforce edition stack. The agent runs against Salesforce data, calls out to integrations, and counts every interaction against a per conversation meter and a per action ledger.

Three pricing pillars

1. Conversation metric. A conversation is the primary billable unit, defined by a 24 hour inactivity window.
2. Einstein Request ledger. Every action the agent takes (search, create record, query, integration call) consumes Einstein Requests.
3. Data Cloud subscription. Most production agents pull from Data Cloud unified profiles, requiring a separate Data Cloud line.

The conversation metric

The conversation metric is the headline meter on every Agentforce quote. The unit price sits at 2 USD per conversation at list, with deep discount only past 250K conversations per month.

Five conversation rules every buyer should know

• 24 hour window. A conversation is bounded by 24 hours of inactivity, not a single message exchange.
• Channel agnostic. A web chat, a WhatsApp thread, and a voice call all count as one conversation each.
• Internal counts the same. An employee facing agent in Service Cloud counts the same as a customer facing one.
• No partial credit. A one message conversation counts as one full unit.
• Pool, not per channel. The conversation pool is contracted across all channels.

Conversation discount band by volume

Monthly conversations	List per conversation	Net per conversation at typical discount
Up to 25K	2.00 USD	1.50 to 1.80 USD
25K to 100K	2.00 USD	1.10 to 1.40 USD
100K to 250K	2.00 USD	0.80 to 1.10 USD
250K to 1M	2.00 USD	0.50 to 0.80 USD
1M plus	2.00 USD	0.30 to 0.50 USD

The action consumption ledger

Every action the Agentforce agent takes consumes Einstein Requests. The ledger tracks search, record retrieval, record creation, integration call, and Apex invocation. Each unit is one Einstein Request.

Five actions that drive Einstein Request consumption

1. Topic match. The agent classifies the user query against a Topic, consuming one Einstein Request.
2. Action invocation. Each action the agent runs consumes one Einstein Request.
3. Record retrieval. A SOQL query or a knowledge article retrieval consumes one Einstein Request.
4. External callout. Each MuleSoft, REST, or HTTP callout consumes one Einstein Request.
5. Long context call. A long context LLM call (over 8K tokens) consumes additional Einstein Requests.

The Data Cloud dependency

Most production Agentforce deployments require a Data Cloud subscription. Data Cloud unifies customer data across Salesforce clouds and external sources into a Customer 360 Profile.

Three Data Cloud rules every Agentforce buyer should know

• Profile metric. Data Cloud prices on the Customer 360 Profile count, not on user count.
• Storage and processing add ons. The base subscription excludes some processing volume.
• Bundle math. Some Agentforce bundles include a Data Cloud starter pack, but only enough for proof of concept volumes.

Indicative Data Cloud cost layered on Agentforce

Profile count	Data Cloud net annual	% of Agentforce run rate
250K profiles	120K to 180K USD	5 to 8 percent
1M profiles	320K to 450K USD	8 to 12 percent
5M profiles	900K to 1.4M USD	12 to 18 percent
20M plus profiles	2.5M to 4M USD	15 to 22 percent

Bundle vs standalone math

Salesforce sells Agentforce both as a standalone add on and inside Service Cloud Einstein for Service bundles. The bundle math hides the per conversation unit price.

Three bundle traps to anticipate

1. Bundle conversation pool too low. A bundle that ships 50K conversations per month against a 200K forecast burns the contract in three months.
2. Bundle Data Cloud starter. Most bundles ship a starter Data Cloud profile pack, not the full production subscription.
3. Bundle Einstein Requests capped. The Einstein Request ledger is also capped inside the bundle.

Worked example: a 250K conversation per month estate

The example below runs a 250K conversation per month forecast through the conversation pool, the action ledger, and the Data Cloud dependency.

250K conversation per month run rate

Component	Volume	Net unit	Annual run rate
Conversation pool	3M conversations per year	0.95 USD	2.85M USD
Einstein Requests on actions	30M requests per year (10 per convo)	0.06 USD	1.80M USD
Data Cloud at 1M profiles	1M profiles	0.38 USD per profile per year	0.38M USD
Total Agentforce annual			5.03M USD

Seven Agentforce negotiation levers on a renewal

Seven levers procurement carries to the table

1. Conversation pool. Negotiate a flat annual pool, not a per Topic cap.
2. Action ledger discount. Discount the Einstein Request ledger separately from the conversation pool.
3. Data Cloud unbundling. Quote Data Cloud as a standalone subscription, not bundled.
4. Overage pricing. Lock the overage at the same net unit price as the contracted pool.
5. Topic scope governance. Limit production Topics to a documented business process list.
6. Channel scope. Lock the conversation pool across all channels with no per channel sub cap.
7. Annual uplift cap. Lock the renewal uplift at zero or a published index.

What to do next

The eight step checklist takes an Agentforce deployment from a sales pitch to a buyer side renewal position.

1. Forecast the conversation volume by channel and by use case.
2. Score the action ratio per conversation across documented Topics.
3. Inventory the Data Cloud profile count against unified records.
4. Quote the bundle and the standalone math on the same forecast.
5. Pressure test the overage clause against forecast variance.
6. Document the Topic scope against the active business process list.
7. Open the renewal nine months early with the conversation forecast and Topic list in writing.
8. Lock the seven levers in a renewal LOI before the SOW.

Frequently asked questions

How Redress engages on Agentforce licensing

Redress runs Agentforce license advisory inside the Vendor Shield subscription, the Software Spend Assessment, the Renewal Program, and the Benchmark Program.

Read the related Salesforce hub, the Salesforce services page, the Salesforce licensing guide, the Salesforce renewal playbook, the utilization calculator, the GenAI hub, the benchmarking page, the about us page, and the contact page.