Salesforce Agentforce Pricing 2026: The True Cost Breakdown

Salesforce Agentforce costs $2 per conversation. Clean number, easy to budget. At 10,000 monthly support interactions, that's $20,000 before you've added a Salesforce license, touched Data Cloud, or accounted for the 30–40% of sessions where the AI didn't resolve anything but charged you anyway. This article breaks down what Agentforce actually costs at real support volumes, what the $2 doesn't cover, and how per-conversation pricing stacks up against per-resolution and flat-rate models.

What does "per conversation" mean in Salesforce Agentforce?

A "conversation" in Agentforce is a customer interaction session, billed from the moment contact is initiated to the moment the session closes. Resolution, human transfer, customer abandonment: all three end the same way from a billing standpoint. The charge is on the session, not the outcome.

Put it in practical terms. A customer contacts your support team three times about an unresolved order issue. Three conversations. Six dollars in AI charges. The AI can close the ticket on the first try and cost $2, or hand off to a human agent on the first try and cost $2. Same price either way.

For teams running 80%+ AI resolution rates, this doesn't sting much. Pay $2 per session when 80% of sessions turn into resolutions, and the effective cost per resolved ticket is $2.50. Tolerable. But at a 50–60% resolution rate, which is where most teams land in the first few months of deployment, roughly half the per-conversation spend goes toward interactions the AI couldn't finish.

What does Agentforce actually cost at 5,000–20,000 tickets per month?

The math here is simple. At $2 per conversation, you're paying for every contact regardless of what the AI does with it. The table below models total monthly usage cost across common support volumes at a 60% AI resolution rate, the threshold CoSupport AI guarantees by day 60 and a reasonable benchmark for a mature deployment.

The effective per-resolved-ticket cost locks in at $3.33 and stays there regardless of volume. Both the bill and the resolution count scale at the same rate. What changes is the absolute dollar figure, and at 20,000 monthly conversations, you're looking at $40,000 per month in AI usage alone.

Drop the resolution rate and the picture gets worse. At 40% resolution and 10,000 monthly conversations, you're resolving 4,000 tickets and paying $20,000. That's $5.00 per AI-resolved ticket.

What the $2 doesn't cover

The per-conversation charge sits on top of the Salesforce platform. Several costs live underneath it.

Salesforce Service Cloud license: Agentforce for Service requires an active Service Cloud subscription. Service Cloud Professional starts at $75 per user per month. Enterprise is $150 per user per month. Unlimited is $300 per user per month. A 10-agent support team on Enterprise: $18,000 per year before a single AI conversation gets initiated.

Data Cloud entitlements: Agentforce grounds its responses in real customer and product data using Salesforce Data Cloud. Data Cloud is a separate entitlement, bundled in some Unlimited editions and licensed separately in others. It's not in the $2.

Flex Credits: Salesforce sells Agentforce capacity as "Flex Credits," where one conversation consumes 10 credits. You purchase credit packs at volume. This doesn't reduce the per-conversation rate. It changes how you buy capacity. The cost per conversation stays at $2.

Configuration and maintenance: Agentforce works by matching customer intent to Topics and then executing linked Actions. Building those Topics and Actions for your product, policies, and edge cases takes real time. So does maintaining them when your product changes. Teams without in-house Salesforce expertise typically need consulting or professional services to stand this up correctly.

The practical baseline for a mid-size team already on Service Cloud Enterprise: $18,000 or more per year in platform costs, then $2 per conversation on top of that.

What most buyers get wrong about Agentforce pricing

The biggest mistake is treating $2/conversation as the complete cost of AI-powered support. It's an incremental usage charge on an existing platform investment.

Salesforce is an enterprise CRM and service platform. Agentforce is a capability that runs inside it. If your support operation already lives in Salesforce, with agents in Service Cloud, customer data in Salesforce CRM, and workflows built in Flow, then per-conversation pricing is the relevant marginal cost and the platform overhead is already paid for. In that context, $2 per conversation is a reasonable number to model.

For teams evaluating AI support who aren't on Salesforce, the arithmetic is different. You'd be buying into a full enterprise platform to access the AI layer, then paying per conversation on top of it. The $2 headline figure doesn't reflect that combination.

Another common error: comparing per-conversation AI costs against the cost of a human agent handling the same volume. The better comparison is against what you'd pay to achieve the same resolution count using an AI model with different billing. When 40% of your AI conversations still end with a human, you're paying $2 for those interactions and then paying your agents to close them. Both costs are real.

Finally: don't assume the $2 rate holds for the life of your contract. Per-conversation pricing in enterprise software can shift with volume tiers, renewal terms, or bundle restructuring. Before building a multi-year ROI model on Agentforce pricing, verify the current rates with Salesforce directly.

How AI customer support pricing models compare

There are three billing structures in the AI customer support market, each with a different risk profile.

Per-resolution pricing shifts the outcome risk to the vendor. If the AI doesn't close the ticket, you don't pay for that interaction. At $0.19 per resolved ticket, 6,000 resolutions from 10,000 monthly conversations costs $1,140. The same 10,000 conversations under Agentforce's per-conversation model costs $20,000. That's a difference of $18,860 per month, or $226,320 per year. The number is stark but it's just arithmetic.

The flat plan works differently. At $99 per 1,000 monthly tickets, CoSupport AI prices the full contact volume regardless of resolution outcome. At $0.099 per ticket, it comes in below both the per-resolution rate and the per-conversation rate. The trade-off is committing to a volume tier each month and not getting credit if contact volume runs lower. A full breakdown of how these three pricing models compare is worth reading before committing to any structure.

For teams at 5,000–20,000 contacts per month with 50–65% AI resolution rates, per-conversation pricing consistently produces the highest total AI spend. You're paying $2 for every session in the 35–50% the AI handles partially, with no discount for the handoffs. If you're specifically looking at Salesforce Agentforce alternatives, the comparison covers tools that operate outside the Salesforce stack entirely.

Key takeaways

• The $2 headline is real. The total cost is something else.
• Agentforce bills per session, not per resolved ticket. At 60% resolution, you're paying $3.33 for every ticket the AI actually closes.
• The $2 usage charge sits on top of Salesforce Service Cloud licensing, which starts at $75 per user per month. A 10-agent team on Enterprise pays $18,000 per year before Agentforce generates a single response.
• For teams already inside the Salesforce ecosystem, Agentforce is a natural extension and the per-conversation cost is the marginal number that matters.
• For teams not on Salesforce, the total cost includes the full platform investment plus per-conversation usage. AI-native alternatives with per-resolution or flat-rate pricing are substantially cheaper at most support volumes.
• At 10,000 monthly conversations, the annual gap between per-conversation pricing ($240,000) and per-resolution pricing at $0.19 ($13,680 at 60% resolution) is over $226,000. Not a rounding errors.