Einstein AI & Agentforce enterprise pricing.
Salesforce priced its AI platform across four dimensions at once: per-user Einstein add-ons, per-conversation Agentforce consumption, the Einstein 1 bundle, and Data Cloud dependency. This note separates those layers, keeps every published rate, and sets out where the budget risk and the negotiation leverage actually sit.
Agentforce's $2 per conversation headline is the smallest part of the bill. The real cost sits in the Einstein 1 edition upgrade — a 200–300% per-seat jump — and the Data Cloud the agents depend on. Model all four layers before signing, and never accept list: we have negotiated conversation rates below $1.20 and cut total AI spend by more than $3M/year at scale.
01 Key findings
Four cost layers, not one price. Total Salesforce AI cost is the sum of the edition baseline, the Einstein 1 upgrade, per-conversation Agentforce consumption, and Data Cloud credits. Salesforce proposals foreground the $2 conversation rate and obscure the other three.
The edition upgrade dwarfs the conversation bill. Moving 1,000 users from Enterprise Sales ($165) to Einstein 1 Sales ($500) adds $4.02M a year — before a single Agentforce conversation is billed.
Consumption pricing is highly negotiable. The $2 rate is list. We have secured $0.85–$1.20 per conversation for commitments above 500,000 conversations a year on multi-year terms.
Data Cloud is the dependency enterprises discover too late. Agents need unified customer data to be useful; the bundled Data Cloud allocation rarely covers organisations with 5M+ profiles, forcing a separate contract.
Copilot and Agentforce are distinct buys. The user-facing assistant and the autonomous agent platform have different ROI drivers. Unbundling them lets you build and negotiate each case on its merits.
02 The consumption model
Agentforce is Salesforce's autonomous agent platform: digital agents that handle service conversations, qualify leads and resolve tickets end-to-end. The pricing logic breaks from the seat-based model — you pay per conversation the agent handles, not per human user. That aligns spend with automation scale, but introduces cost variability seat licences never had.
The published rate is $2 per conversation. A conversation begins when a user initiates an interaction and ends when it is resolved or handed to a human — multiple turns (greeting, authentication, diagnosis, resolution) count as one conversation, which is more favourable than token-based billing. But the $2 rate covers standard agent types only: custom agents built in Agentforce Studio, extended context windows, or external-API reasoning chains can price differently. Confirm with your AE exactly which scenarios sit inside the standard rate.
The $2 rate is list pricing. We have negotiated Agentforce as low as $0.85–$1.20 per conversation for enterprise commitments above 500,000 conversations per year on multi-year terms. Consumption models are highly negotiable when volume commitments are on the table — never sign at list.
03 Einstein & Agentforce pricing
Most Einstein AI capabilities are bundled into Salesforce's higher-tier platform editions. The Einstein 1 tiers — Sales, Service and Platform — package AI features with CRM. Every published list rate we track is below; the upgrade from Enterprise represents a 200–300% per-seat increase.
| Edition / SKU | List price | Key AI features included |
|---|---|---|
| Enterprise (Sales/Service) | $165 / user / mo | Basic Einstein predictions, Next Best Action |
| Einstein 1 Sales | $500 / user / mo | Einstein Copilot, Deal Insights, Conversation Intelligence, Data Cloud access |
| Einstein 1 Service | $500 / user / mo | Einstein Copilot, Case Classification, Summarisation, Knowledge Grounding, Data Cloud access |
| Einstein 1 Platform | $700 / user / mo | Full Agentforce builder access, custom agent creation, all Einstein 1 features across clouds |
| Agentforce (standard agents) | $2 / conversation | Per-conversation consumption; negotiable to $0.85–$1.20 at volume |
| Einstein Conversation Intelligence (add-on) | ~$50 / user / mo | Call analysis, coaching insights, deal-risk signals — standalone on Enterprise |
Einstein 1 Copilot features — AI-drafted emails, call and case summaries, deal-health scores — are genuinely useful, but Copilot and Agentforce are two distinct value propositions. Isolate them: Conversation Intelligence added standalone (~$50/user/mo) is often cheaper than a full Einstein 1 Sales upgrade if it is the real driver, and Agentforce Studio (custom agent building) needs Einstein 1 Platform at $700/user only for the technical team, not every end user.
04 Cost at enterprise scale
Consider a 1,500-user enterprise (800 Sales, 700 Service on Enterprise) deploying Agentforce for service deflection at 200,000 conversations a month. Incremental annual cost before negotiation runs to roughly $6M–$8M. The layers, at list:
Even a smaller shop feels it: 100,000 service interactions a month with 30% deflected to agents is $60,000/month — $720,000 a year — in Agentforce consumption alone, before licensing or Data Cloud. At the 1,500-user scale, the gap between accepting list and securing negotiated rates can exceed $3M per year.
05 The consumption-overrun trap
The dimension enterprises discover too late is Data Cloud. Agents need a unified customer data layer — purchase history, case history, contract status, recent interactions — to give contextually intelligent answers rather than generic, rules-based responses. Einstein 1 editions bundle a starter Data Cloud allocation, but organisations with 5M+ profiles, real-time event streams or multiple sources will need a dedicated Data Cloud contract to support production Agentforce. See our Data Cloud pricing guide for the credit-consumption detail.
Consumption compounds where you cannot see it. Agentforce proposals emphasise the per-conversation rate while the edition upgrade and Data Cloud credit burn accumulate underneath. Every Copilot interaction that touches a Data Cloud profile consumes query credits; every uncapped deflection scenario scales the conversation bill. Model all three variable layers together, and cap them contractually — or the "$2" number becomes the least of your worries.
06 Cost-driver framework
Four variables decide what you actually spend. Weight them to your deployment before committing to any Einstein 1 or Agentforce contract.
Conversation volume & deflection
Model deflection rates and monthly conversation volumes conservatively. This single input drives Agentforce consumption — and the volume commitment that unlocks sub-$1.20 rates.
Edition-upgrade scope
Not every user needs Einstein 1. Assign Platform ($700) to agent-builders only and keep end users on lower tiers. The upgrade, not the conversation, is the biggest line item.
Data Cloud sizing
Profile count, data sources and real-time streams determine whether the bundled allocation holds or a dedicated contract is required. Size it before, not after, go-live.
Pilot before commitment
A structured 90-day pilot on real interactions produces the consumption data that makes your enterprise volume model — and your negotiating position — defensible.
07 Negotiation playbook
AEs bundle Copilot with Agentforce in one Einstein 1 proposal. Split them, build independent business cases, and negotiate each on its merits rather than a package discount that may not match your usage.
Consumption pricing is volume-sensitive. Offer a commitment 20–30% below your central estimate — enough to capture the volume discount, with downside protection if adoption ramps slower than projected.
Force a 90-day pilot at reduced rates before enterprise volumes, and attach performance SLAs — deflection, classification accuracy, fee credits. See our AI contract clauses guide.
Negotiating an Agentforce or Einstein deal?
Our vendor-negotiation practice models total AI platform cost and secures the rate — before you sign, not after.
The Licensing Edge
Weekly vendor intelligence on Salesforce AI, Microsoft Copilot and enterprise software. 3,000+ IT leaders subscribe.