Salesforce Agentforce Licensing Guide:
AI Agents, Pricing Models & Negotiation Strategies

What Agentforce Actually Is

Agentforce is Salesforce's agentic AI platform — purpose-built autonomous agents that can handle customer conversations, qualify leads, resolve service cases, and execute workflows without human handoff. It launched at Dreamforce 2024 as the successor to Einstein Copilot, and by early 2026 had become the central commercial focus of every Salesforce renewal conversation.

What makes Agentforce commercially distinct is its pricing model. For the first time in Salesforce's history, the platform introduced outcome-based, consumption-led billing alongside its traditional per-seat structure. This shift has created significant confusion — and significant overspend — for buyers who did not fully model the cost before signing.

The Three Pricing Models

Salesforce offers three commercially distinct ways to buy Agentforce. Understanding which model applies to your use case — and the trade-offs between them — is the starting point for any accurate cost model.

Model 1 — Conversation-Based ($2 per conversation)

The original Agentforce pricing. A conversation begins when an agent first responds to a user and ends when the user closes the session, escalates to a human, or the conversation goes inactive for 24 hours. All exchanges within that window count as one conversation, regardless of how many turns occur. Best suited to customer-facing public-website bots with predictable, bounded interactions. Volume discounts are available for high-engagement commitments.

The problem with $2/conversation emerged quickly in enterprise deployments: a single complex service conversation with 80 back-and-forth exchanges involving case lookups, knowledge retrieval, and record updates could incur $2 in conversation cost plus meaningful Einstein Request and Data Cloud credit consumption on top. For variable, action-heavy enterprise workflows, this model was unpredictable and often more expensive than it appeared.

Model 2 — Flex Credits ($0.10 per action / 20 credits per action)

Introduced in May 2025 in direct response to enterprise backlash against the conversation model. Flex Credits are a pay-per-action currency: each discrete task an AI agent performs — updating a record, summarising a case, scheduling an appointment, answering from a knowledge base — consumes 20 Flex Credits. Credits are priced at $500 per 100,000 credits, making each action cost $0.10 at list rate. Credits cannot be combined with the conversation model in the same org — you choose one model per deployment.

Flex Credits are fungible across teams, channels, and use cases, and can be pooled across contracts in some configurations. Unused credits expire at the end of the subscription term — a meaningful risk for buyers who overestimate consumption in their first year.

Model 3 — Per-User Add-ons and Editions

For internal, employee-facing use cases where predictable flat-rate cost is preferred over consumption billing, Salesforce offers per-user add-ons starting at $125 per user per month for Sales, Service, and Field Service — providing unlimited internal agent usage for licensed users. Industry-specific add-ons (Financial Services, Healthcare, Manufacturing) are priced at approximately $150 per user per month, reflecting the compliance and workflow overhead of regulated-sector deployments. Agentforce 1 Editions bundle the add-on with large pools of Flex Credits and Data Cloud capacity at approximately $550 per user per month — a significant per-user cost that is Salesforce's attempt to make AI adoption a wholesale platform commitment rather than a consumption add-on.

Flex Credits: How the Maths Works

Flex Credits are simple in principle but require careful modelling in practice. Real-world consumption examples from Salesforce's own documentation:

• Case management workflow — 3 actions (identify customer, retrieve cases, add comment): 60 credits = $0.30 per case. At 100 users managing 3 cases/day, 20 days/month: 360,000 credits/month = $1,800/month in Flex Credit cost alone, before any Data Cloud or Einstein Request costs.
• Employee onboarding query — 1 action (knowledge base search): 20 credits = $0.10 per query. At 20 new employees/month each asking 5 questions: 2,000 credits = $10/month. Very low volume, very economical.
• Field service scheduling — 6 actions (time preference check, availability lookup, engineer matching, confirmation): 120 credits = $0.60 per appointment.

The maths scales linearly — but Agentforce workflows in production often involve more actions than the simplified examples suggest. A customer-facing service agent handling a complex billing dispute might execute 8–12 actions (record lookups, policy checks, eligibility verification, case creation, notification sends). At 120–240 credits per interaction, the per-interaction cost of $0.60–$1.20 begins to approach the $2 conversation rate. Model your most complex use cases at realistic action counts before committing to Flex Credits.

Hidden Dependencies That Drive Real Cost

Agentforce does not operate in isolation. Three cost layers sit beneath the headline Agentforce pricing that buyers routinely miss:

Einstein Requests

AI-generated content within conversations (responses, summaries, generated text) consumes Einstein Requests — measured in 1,500-word units. A long, complex AI-generated response can consume 10–15 Einstein Requests. For orgs using prompt templates outside Agentforce (via Flow, Apex, or LWC), Einstein Requests — not Flex Credits — are consumed. These are billed separately and can add $2–$3 per 1,000–1,500 words of AI-generated output at market rates.

Data Cloud Credits

Agentforce agents that retrieve customer context, perform identity resolution, or activate data segments draw on Data Cloud credits. Every Agentforce deployment that uses personalised, data-grounded responses — which is most production use cases — will consume Data Cloud credits in addition to Agentforce conversation or Flex Credit costs. These are billed separately through the Data Cloud consumption model.

Enterprise Edition Requirement

Agentforce requires Salesforce Enterprise Edition or higher as the base platform. Organisations on Professional Edition must upgrade their base platform before Agentforce can be deployed — adding $75–$100 per user per month in platform uplift before the first Agentforce credit is consumed.

Add-ons vs Agentforce 1 Editions

The choice between Agentforce add-ons and the Agentforce 1 Edition bundle is the most consequential commercial decision in the buying process:

Agentforce 1 Edition at $550/user/month is Salesforce's full-platform AI commitment — equivalent to a new top-tier Salesforce licence. Before signing, model whether the included credit pools will actually be consumed at your projected usage levels. Buying a large credit bundle that expires unused is not value; it is pre-paid waste.

Negotiation Strategies

• Run both pricing models before committing. Map your 5–10 highest-volume Agentforce use cases, calculate their action counts, and price them under both $2/conversation and Flex Credits. The model that wins depends on your interaction complexity — not on Salesforce's preference.
• Negotiate credit rollover. Unused Flex Credits expire at term end by default. Push for at least a one-year rollover on unused credits — particularly valuable in year one when adoption typically ramps slower than projected.
• Pre-agree overage rate before committing. If you commit to a Flex Credit volume and exceed it, Salesforce will charge overage at a rate you should negotiate upfront. Without a pre-agreed rate, you are billed at list. Lock in the same per-credit rate as your initial commitment for overages.
• Tie Agentforce 1 Edition to measurable adoption milestones. If Salesforce is pushing Agentforce 1 Edition, negotiate the right to scale back to add-on pricing at renewal if AI adoption metrics — agent conversations handled, automation rate, case deflection — do not hit agreed targets.
• Use early-adopter leverage. Salesforce is aggressively incentivising Agentforce adoption. Early-commitment buyers can negotiate deeper discounts on Flex Credit bundles, Data Cloud credits, and platform uplift — but only before signing, not after.

Pre-Signature Checklist

• Pricing model chosen — conversation vs Flex Credits modelled against your actual use cases
• Credit rollover negotiated — unused credits carry forward at least one year
• Overage rate pre-agreed — same per-credit rate as committed volume, confirmed in contract
• Einstein Request cost modelled — AI-generated output volume estimated and priced separately
• Data Cloud dependency confirmed — Data Cloud credit requirements scoped for grounding and personalisation use cases
• Base platform tier confirmed — Enterprise Edition requirement validated; uplift cost included in total
• Adoption milestone gates in multi-year contracts — right to reduce scope if usage targets not met

Related Guides