Contents
Agentforce represents Salesforce's full pivot to autonomous AI — moving beyond Einstein's predictive analytics and Copilot's generative assistance to agents that independently take actions: updating records, executing workflows, responding to customers, and escalating to humans only when necessary. Launched in late 2024 and expanding rapidly through 2025 and 2026, it is now a fixture of nearly every Salesforce enterprise renewal conversation.
The pricing model — $2.00 per conversation at list price — sounds simple. It is not. Understanding exactly what constitutes a billable conversation, what is included in your current platform, and how to negotiate usage-based pricing before you have usage data is a challenge that requires both commercial experience and technical understanding.
What Is Agentforce?
Agentforce is Salesforce's platform for deploying AI agents across customer service, sales, marketing, and internal IT processes. It builds on the Salesforce platform's existing data model, flows, and integration architecture, which means agents can access CRM records, execute business processes, and interact with customers via any channel (chat, email, voice, messaging).
Key agent types available in 2026 include:
- Service Agent: Handles customer enquiries autonomously, resolves cases using knowledge base content, and escalates to human agents when confidence thresholds are not met
- Sales Development Representative Agent: Qualifies inbound leads, books meetings, sends personalised follow-up sequences
- Marketing Agent: Personalises campaign content, responds to prospect enquiries from marketing automation
- Commerce Agent: Assists customers through e-commerce journeys, handles returns and order management
- Custom Agents (Agent Builder): Configurable agents built using Agent Builder for any enterprise use case
The underlying intelligence is powered by Salesforce's Hyperforce infrastructure, large language models (primarily accessed via the Einstein Trust Layer, which routes requests to models from multiple providers including Anthropic and Google without persistent training on customer data), and the Atlas Reasoning Engine — Salesforce's proprietary orchestration layer that breaks tasks into sub-steps and executes them autonomously.
The Per-Conversation Pricing Model Explained
Agentforce is priced at $2.00 per conversation at published list price. A "conversation" in Salesforce's definition is an interaction where an Agentforce agent handles a customer or user request — performing at least one action before either resolving the request autonomously or transferring to a human agent.
The definition of "conversation" is not uniformly enforced and has been interpreted inconsistently across early Agentforce deployments. Before signing, require Salesforce to provide a written, legally binding definition of what constitutes a billable conversation — including whether failed escalations, test interactions, and internal agent-to-agent orchestration steps count toward the billing total.
Enterprise volume pricing substantially reduces the per-conversation cost:
| Annual Conversation Volume | Per-Conversation Rate (Est.) | Annual Cost Estimate |
|---|---|---|
| Up to 10,000 | $2.00 (list) | Up to $20,000 |
| 10,000–100,000 | $1.40–$1.75 | $14,000–$175,000 |
| 100,000–500,000 | $1.00–$1.40 | $100,000–$700,000 |
| 500,000–2M | $0.70–$1.00 | $350,000–$2M |
| 2M+ | Negotiated individually | Custom |
These are enterprise-negotiated estimates based on our engagements. Actual rates vary significantly based on your overall Salesforce contract size, renewal timing, and competitive leverage. Organisations spending $1M+ annually with Salesforce have substantially more pricing power on Agentforce than those spending under $200K.
Forecasting Your Conversation Volume
One of the fundamental challenges of Agentforce pricing is that you are committing to a usage-based model before you have production usage data. Salesforce's sales teams will encourage you to commit to volume packs at discounted rates — but the incentive structure means they may propose volumes that benefit them, not you.
Our recommended approach is to:
- Pilot first, commit second: Run a 90-day paid pilot at a low conversation commitment before signing a multi-year volume pack
- Model based on human agent volumes: If Agentforce is replacing human interactions, your existing contact centre data (monthly interactions, resolution rates, escalation rates) is your best volume proxy
- Build in overage protections: Negotiate a ceiling on per-conversation overage rates — ideally no more than 10–15% above your committed rate, not list price
- Negotiate rollover credits: Unused conversation credits from one year should carry forward, not expire — this is a key contractual protection
Agentforce in Einstein 1 Bundles
If you are on an Einstein 1 edition of Sales Cloud or Service Cloud, you receive an Agentforce "Foundations" entitlement — a limited conversation allocation (typically 200–500 conversations per user seat per year, though allocations vary) included in your platform fee. This is meaningful for smaller-scale deployments but typically insufficient for any enterprise deploying Agentforce at the contact centre or revenue-generation level.
Salesforce's commercial team will frequently use Agentforce as justification for upgrading customers from standard editions to Einstein 1. Before agreeing to an edition upgrade, calculate whether the per-conversation pricing from Agentforce Foundations inclusions actually justifies the edition premium. In many cases, it does not — and you are better off on a standard Enterprise edition with a standalone Agentforce volume purchase.
Salesforce's internal guidance as of early 2026 is to drive Einstein 1 adoption using Agentforce as the primary value justification. Account teams receive higher accelerators on Einstein 1 upgrades than on standalone Agentforce purchases. Knowing this allows you to negotiate the Einstein 1 upgrade price down substantially — or to decline the upgrade while securing favourable standalone Agentforce pricing.
Contract Terms and Risks to Address
Agentforce contracts introduce a new category of risk that did not exist in traditional SaaS agreements. The following terms require explicit attention:
Conversation Definition and Audit Rights
Insist on a written, enforceable definition of a billable conversation. Include your right to audit Salesforce's conversation count methodology, including access to logs that demonstrate how each billed conversation was determined. This is standard practice in usage-based software contracts; Salesforce should agree to it.
Price Stability Across the Contract Term
Lock in your per-conversation rate for the duration of your contract — 2 or 3 years. Agentforce pricing is evolving rapidly; Salesforce may seek to reprice at renewal. Securing a multi-year rate today protects you from pricing risk as the product matures and Salesforce gains confidence in its value metrics.
Agent Configuration and IP Ownership
The agent configurations you build in Agent Builder — system prompts, action sequences, knowledge base integrations — represent meaningful intellectual property. Ensure your contract explicitly states that all agent configurations are owned by your organisation, not Salesforce, and that you retain the right to export them on contract termination. See our broader guide on AI IP ownership in vendor contracts for the full clause framework.
Data Processing and Training
Salesforce's Einstein Trust Layer is designed to prevent customer data from being used to train Salesforce's foundation models. However, your contract should explicitly confirm this, specifying: (a) that conversation data is not used for any model training, (b) where conversation logs are stored and for how long, and (c) your right to delete conversation data on request.
Service Level Agreements
Standard Salesforce SLAs cover platform uptime. Agentforce introduces a new category of SLA: agent response accuracy, resolution rate, and escalation correctness. These are not currently included in standard contracts but should be addressed in your pilot terms and, if the deployment is business-critical, in your production contract.
Data Governance and IP Ownership
Agentforce agents necessarily access your Salesforce data to function — customer records, case history, product data, pricing rules. This creates data governance considerations beyond the standard Salesforce data processing agreement:
- Access controls: Which data objects can each agent access? Ensure Agent Builder's permission model is configured with least-privilege principles, and that these controls are contractually documented.
- Conversation logs: Where are conversation transcripts stored? In your Salesforce org? On Salesforce infrastructure outside your org? Clarity here is required for GDPR, CCPA, and sector-specific compliance.
- Third-party model routing: Salesforce's Einstein Trust Layer routes some requests to third-party models (Anthropic Claude, Google Gemini, others). Confirm in writing which models may process your data and under what terms.
- Right to audit: Retain the right to audit Salesforce's data processing practices related to Agentforce, particularly if you operate in regulated industries (financial services, healthcare, government).
For a comprehensive checklist of AI contract governance requirements, see our guide on AI governance in vendor contracts and our white paper on AI Vendor Contract Red Flags.
Negotiation Strategy for Agentforce
Agentforce negotiations in 2026 benefit from the fact that Salesforce is aggressively pushing adoption — which means commercial flexibility is higher than it will be once the product is established. This window will not remain open indefinitely. Key tactics:
1. Pilot Before Committing to Volume
Negotiate a capped pilot of 3–6 months at a fixed conversation budget. Use the pilot data to forecast annual volume with confidence before committing to a multi-year volume pack. Salesforce will apply pressure to skip this step — resist it.
2. Use Agentforce Adoption as Negotiation Currency
Salesforce has internal adoption targets for Agentforce. Your willingness to deploy and publicise Agentforce (case study, reference call) has genuine commercial value to Salesforce. Use this as leverage for pricing, included services (implementation support), and contractual protections.
3. Benchmark Against Competitive Agent Platforms
ServiceNow AI Agents, Microsoft Copilot Studio, and Genesys Cloud CX all offer comparable agent capabilities for customer service use cases. Request demos and pricing — visibly — from at least two competitors. This shifts Salesforce's commercial posture immediately.
4. Negotiate the Full Bundle, Not Just Agentforce
If Agentforce is being layered onto a Salesforce renewal, negotiate the entire contract — platform license, add-ons, and Agentforce — as a single transaction. The aggregate value of the deal gives you maximum discount authority.
Navigating Your Agentforce Negotiation?
Our former Salesforce executives have been inside these negotiations from both sides. We identify the terms that matter and the leverage points that move pricing.
Request a Confidential Briefing Download AI Contract Red Flags GuideCompetitive Alternatives
The following platforms compete with Agentforce for enterprise AI agent deployments and serve as credible alternatives for negotiation leverage:
- Microsoft Copilot Studio: Deep Azure and Microsoft 365 integration, per-message pricing model, strong for internal-facing agent deployments. Best leverage if you are a large Microsoft EA customer.
- ServiceNow AI Agents: Purpose-built for IT service management and employee experience. Competitive for organisations deploying agents in ITSM and HR service delivery contexts.
- Genesys Cloud CX: Enterprise contact centre platform with AI agent capabilities. Strongest alternative for customer service deployments where contact centre integration is critical.
- Anthropic Claude via API: For organisations with engineering capability, building custom agents on Claude's API can achieve comparable outcomes at significantly lower per-interaction costs. See our guide on Anthropic Claude enterprise pricing.
Related Resources
- The Complete Guide to Salesforce Contract Negotiation
- Salesforce Einstein AI Pricing Guide
- Salesforce Marketing Cloud Pricing Guide
- AI IP Ownership: Contract Clauses That Protect You
- AI Governance in Vendor Contracts
- AI Procurement Advisory Service
- AI Vendor Contract Red Flags (White Paper)
Frequently Asked Questions
How much does Salesforce Agentforce cost?
Salesforce Agentforce is priced at $2.00 per conversation at list price. Enterprise accounts purchasing volume packs of 100,000+ conversations typically negotiate rates of $0.80–$1.40 per conversation. Conversations are also included in Einstein 1 editions at a set allocation per user seat.
What is a "conversation" in Agentforce pricing?
A conversation is an interaction where an Agentforce agent handles a request and takes at least one action — retrieving data, executing a flow, or generating a substantive response — before resolving or escalating. The definition requires explicit contractual clarification; ensure it is defined in writing before signing.
Is Agentforce included in existing Salesforce Einstein licenses?
Agentforce Foundations (limited conversation volume) is included in Einstein 1 editions of Sales and Service Cloud. Standard Enterprise, Professional, and Growth editions require a separate Agentforce purchase. Einstein 1 upgrades are being aggressively sold using Agentforce as justification — evaluate whether the upgrade cost is justified by the included conversation allocation.
What contract terms should enterprises negotiate for Agentforce?
Key terms include: written definition of billable conversation, audit rights on usage counting, multi-year price lock on per-conversation rate, rollover of unused credits, IP ownership of agent configurations, data processing limitations (no training use), and performance SLAs for agent resolution accuracy.