Most pricing pages give you the headline number. This guide breaks down the full cost — licences, conversations, Data Cloud, Einstein, implementation, and ongoing support — so you can model a realistic business case before your first scoping call.
Agentforce is priced per conversation — not per seat, not per month. As of 2026, the list price is $2 per conversation with volume discounts negotiated at enterprise level. This is a structural departure from the Salesforce licence model you may be used to: your bill is determined by how much your agents do, not by how many users you have.
Understanding what counts as a conversation is the first pricing question every buyer should resolve before modelling costs.
A conversation is a complete interaction between an end user and an Agentforce agent — from the first message to resolution or escalation. A conversation ends when the user session closes, the agent resolves the case, or a 24-hour inactivity window passes. Multiple back-and-forth messages within one session count as one conversation. A complex exchange involving 15 messages costs the same as a single-question resolution.
This matters enormously for your cost model. If your target use case involves lengthy multi-turn exchanges (complex technical support, financial advisory sessions), the per-conversation unit economics look very different from a simple FAQ deflection use case where most sessions resolve in 2–3 exchanges.
The $2/conversation figure covers your Agentforce conversation consumption. It does not cover the additional licences and platform costs that most serious production deployments require. Here is what you need to budget for beyond the headline number.
Data Cloud is not included in the Agentforce conversation price. If you want your agents grounded in unified customer data — which is the recommendation for any production-grade deployment — you need Data Cloud. Without it, your agents reason over whatever data is in the immediate Salesforce record context. With it, they have access to a unified 360-degree customer profile combining CRM data, engagement history, external signals, and calculated insights.
Data Cloud pricing as of 2026: $108,000/year for 10 million unified profiles (Salesforce list price). Many organisations that have already invested in Salesforce at scale already have Data Cloud — in which case this is not an incremental cost. If you do not have it, this is the largest add-on cost in the stack.
Some Agentforce capabilities — Einstein Copilot Search, Einstein for Sales, Einstein for Service — require Einstein licences on top of your base Salesforce edition. These are user-based licences that sit alongside the conversation-based Agentforce pricing.
Einstein for Sales and Service: approximately $50/user/month. A 50-person sales team adds $30,000/year. If you are on Salesforce Unlimited or Unlimited+, some Einstein features are already included in your base licence — confirm with your AE which specific capabilities are covered before purchasing add-ons.
If your agents need to connect to non-Salesforce systems — ERP for inventory and order data, EHR for patient records, core banking platforms, logistics tracking systems — you need an integration layer. MuleSoft is the Salesforce-native option.
MuleSoft Anypoint Platform: typically $150,000–$500,000+/year depending on cores and API call volume. This is a significant budget line. However, for simpler integrations, Agentforce supports HTTP callout Actions via Apex that can connect to REST APIs directly without MuleSoft — a substantial cost saving for architectures where MuleSoft's advanced transformation and governance capabilities are not needed.
The scoping question to ask early: does every external integration need MuleSoft, or can some be handled via Apex callouts? In our experience, approximately 40% of integrations that clients initially think need MuleSoft can be addressed with Apex — saving six figures annually.
Agentforce runs on top of your existing Salesforce org. Sales Cloud, Service Cloud, Financial Services Cloud, or Health Cloud licences are prerequisites. If you do not have these, they need to be factored into total cost. Sales Cloud Professional: from $80/user/month. Service Cloud: from $80/user/month. Enterprise editions (required for many Agentforce capabilities): from $165/user/month.
For most organisations, implementation is the largest one-time cost and the factor most likely to determine whether the deployment succeeds. Implementation cost is highly variable and depends on use case complexity, org maturity, integration requirements, and the partner model you choose.
One use case, scoped to production. Includes Topic configuration, Action development (Flow and Apex), Einstein Trust Layer setup, human-in-the-loop exception handling, UAT, and go-live. Also includes 2 weeks of post-launch support to tune behaviour and resolve edge cases surfaced by real traffic. This is how Kovil AI always starts — a focused pilot on the highest-value use case, with a clear definition of success.
Multiple use cases across Sales Cloud, Service Cloud, and/or Marketing Cloud. Includes MuleSoft integration (if required), Data Cloud setup, custom Action development, full UAT, training, and production deployment. The upper end of this range reflects complex integrations with legacy systems, regulated industry compliance configuration, and large knowledge base setup for RAG-grounded agents.
Agent performance monitoring, conversation analysis (identifying gaps in Topic coverage), model and instruction updates, new use case addition, and quarterly business reviews. Most organisations find that agents require ongoing tuning in the first 6 months as real-world traffic surfaces edge cases not seen in UAT. The question is whether your internal team handles this or an external partner does.
Here is a worked Year 1 TCO example for a mid-size B2B SaaS company deploying Agentforce for SDR qualification and service case resolution.
Scenario: 5,000 conversations/month, 2 use cases (SDR qualification agent + Service case resolution agent). No existing Data Cloud. 50-user sales team requiring Einstein for Sales licences. 6-week implementation scoped as pilot-then-full-build.
ROI crossover: At 68% case resolution automation, a team that previously needed 10 service agents at $55,000/year loaded cost ($550,000/year) now handles the same volume with 3–4 agents ($165,000–$220,000/year). Gross annual saving: $330,000–$385,000. Net of Year 2 Agentforce cost ($234,000): approximately $96,000–$151,000 net annual saving. Break-even on the Year 1 investment lands at approximately 21–24 months.
Agentforce pricing is not fixed — it is negotiated. The headline $2/conversation figure is the Salesforce list price, not the price sophisticated buyers pay. Here is where the negotiation leverage exists.
Volume commitments: Commit to higher monthly conversation volume and negotiate the per-conversation rate down. If your use case justifies 1,000+ conversations/day, use that volume commitment as your primary negotiating lever.
Multi-year deals: A 3-year commitment typically unlocks better per-conversation pricing and potentially favourable terms on Data Cloud. If you are confident in the platform, the multi-year discount can be worth it.
Bundled Salesforce deals: If you are negotiating or renewing other Salesforce products simultaneously (Sales Cloud, Service Cloud, Marketing Cloud), bundle Agentforce into the negotiation. Salesforce AEs have more flexibility on package pricing than on individual product pricing.
Proof-of-value pilot terms: Ask for a limited-volume pilot at reduced or fixed price before committing to annual volume. Many AEs can accommodate a 90-day, capped-volume pilot arrangement. This lets you validate the business case with real traffic before the annual commitment.
Kovil AI advises on licence optimisation as part of our Strategy & Readiness engagement. We have visibility across multiple deployments and can advise on what realistic negotiated pricing looks like for your volume profile before you walk into the AE conversation.
When you sit down with your Salesforce AE to discuss Agentforce pricing, these are the eight questions that will reveal the true cost and the available flexibility:
What is the all-in per-conversation price for our expected monthly volume, including any commit discount?
What licence tiers are required for the specific use cases we are targeting (service deflection, SDR qualification, internal operations)?
Is Data Cloud included in any packaging option at our Salesforce spend level, or is it a full add-on?
What Einstein capabilities are included in our current Salesforce edition vs. requiring additional licences?
Can we run a 90-day capped pilot at a fixed or trial rate before committing to annual conversation volume?
What is the contractual commitment if our actual conversation volume is significantly lower than our committed volume?
What is included in Agentforce platform support at our licence tier, and what costs extra?
If we bundle Agentforce with our upcoming Sales Cloud or Service Cloud renewal, what is the combined discount available?
Book a scoping call — we will model the full licence and implementation cost for your use case before you commit to anything. No obligation, no sales pitch.
Agentforce is priced per conversation, not per user or per seat. As of 2026, the list price is $2.00 per conversation. This is a fundamentally different model from traditional Salesforce licences — your cost scales with agent activity, not headcount. Volume discounts are negotiated at enterprise level and can bring the effective rate down to $0.80–$1.20 per conversation at scale.
A conversation is a complete interaction between an end user and an Agentforce agent — from the first message to resolution or escalation. A conversation ends when the user session closes, the agent resolves the case, or a 24-hour inactivity window passes. Multiple back-and-forth messages within one session count as a single conversation. One exchange of 12 messages costs the same as a single-message interaction that resolves immediately.
Not strictly required for basic deployments, but strongly recommended for production-grade agents. Without Data Cloud, your agents are grounded only in data available directly in your Salesforce org. With Data Cloud, agents access unified customer profiles that combine CRM data, engagement history, and third-party signals. The cost difference — Data Cloud lists at $108,000/year for 10M profiles — needs to be weighed against the improvement in agent accuracy and resolution rate. Many organisations already have Data Cloud, making this a moot cost question.
MuleSoft is not required. For simpler integrations with external systems, Agentforce can use HTTP callout Actions via Apex — connecting to REST APIs without MuleSoft. MuleSoft becomes worth the cost when you need complex, high-volume integration with enterprise systems (ERP, EHR, core banking) or when you need advanced transformation, monitoring, and governance capabilities. Typical MuleSoft Anypoint Platform pricing ranges from $150,000–$500,000+/year. For lighter integration needs, Apex HTTP callouts are a significant cost saving.
Yes. Salesforce offers monthly Agentforce licences (at a premium over annual pricing), and Kovil AI runs fixed-price pilots scoped to a single use case. Our typical pilot is 2–3 weeks and scoped to production — not a proof of concept. The pilot gives you real conversation volume, real resolution rate data, and a real business case before you commit to annual volume. We recommend starting with a pilot for exactly this reason.
Implementation cost varies by scope. A focused single-use-case pilot (2–3 weeks) typically runs $15,000–$35,000 fixed price. A full multi-cloud deployment (6–12 weeks) covering multiple use cases, MuleSoft integration, and Data Cloud setup runs $60,000–$180,000. Ongoing managed services for agent monitoring and optimisation add $5,000–$15,000/month. Traditional Salesforce SIs often charge $250–$400/hour on T&M, which can reach $432,000+ for a 12-week engagement. Kovil AI operates fixed-price only.
Agentforce is priced per conversation for autonomous agent interactions. Einstein licences are user-based add-ons (~$50/user/month for Einstein for Sales or Service) that unlock Einstein Copilot features, predictive scoring, and generative AI capabilities for individual reps. Many Agentforce deployments need both: Agentforce for autonomous agent handling, Einstein licences for the rep-facing AI features that support HITL escalations. If you are on Salesforce Unlimited or Unlimited+, some Einstein features are included.
Start with your baseline cost: number of service reps or SDRs handling the target query type, multiplied by annual loaded cost per head. Then establish your autonomous resolution rate target (typically 55–70% for well-configured agents). Apply that rate to calculate labour displacement. Compare annual Agentforce cost (conversations × $2 × 12 months, plus licences) against annual labour saving. Typical break-even is 14–24 months. Our ROI guide covers this in full detail with worked examples from real deployments.
Cost savings model, revenue uplift calculation, time-to-ROI analysis, and real deployment numbers from production.
14 min readTechnical GuideAtlas Reasoning Engine, Topics and Actions, Einstein Trust Layer, and how agents take action in external systems.
12 min read