The honest answer: it depends — but here are the numbers
Agentforce pricing has evolved significantly since its launch. As of 2025, Salesforce has moved to a consumption-based model for most Agentforce products, which means your costs scale with usage rather than seats. This is good news for companies starting small, and important to understand before you budget.
Here is the breakdown of what you will actually pay.
Agentforce licensing costs
Agentforce for Service
The core Service Cloud Agentforce product is priced at $2 per conversation. A "conversation" is defined as a complete customer interaction handled by the agent — from first message to resolution or handoff.
What this means in practice: if your service team handles 10,000 cases per month and the agent deflects 60% of them, you are paying for approximately 6,000 agent conversations — roughly $12,000 per month in pure licensing. Compare that to the fully-loaded cost of human agents handling those same 6,000 cases, which at $18-24 per case runs $108,000-$144,000.
Agentforce for Sales
Sales-focused agents — covering SDR outreach, meeting scheduling, lead qualification, and pipeline research — are also consumption-based. Pricing is typically bundled with Sales Cloud Einstein and varies by the specific actions the agent performs.
Most Sales Cloud Agentforce deployments we see run $3,000-$8,000 per month in licensing for a team of 20-50 reps, depending on agent utilization.
Einstein Platform (if building custom agents)
If you want to build fully custom agents beyond the out-of-the-box Service and Sales templates, you will need Einstein Platform licenses. These start at approximately $75 per user per month and give your developers access to Agent Builder, the full action library, and Data Cloud integration.
What drives costs up
Data Cloud
Agentforce performs significantly better when grounded on Data Cloud — it gives agents access to unified customer data beyond what is in standard CRM objects. Data Cloud is licensed separately, starting at around $108,000 per year for the base tier. For many mid-market companies, Data Cloud is optional for a first deployment but becomes important at scale.
Third-party integrations
If your agents need to take actions outside Salesforce — querying your ERP, reading from your data warehouse, writing to external systems — you will need MuleSoft or custom API integrations. This adds implementation cost but not necessarily ongoing licensing cost.
Conversation volume spikes
Because pricing is consumption-based, unexpected volume spikes directly impact your bill. We always recommend setting consumption alerts in Salesforce and building a cost model with a realistic range — base case, upside case, and spike scenario.
Implementation costs
Licensing is only part of the picture. Implementation costs for a first Agentforce deployment typically break down as follows:
| Engagement type | Scope | Typical cost range |
|---|---|---|
| Discovery & strategy only | Use case prioritization, ROI model, architecture blueprint | $15K–$30K |
| Single-agent deployment | One agent, one channel, full implementation | $60K–$120K |
| Multi-agent program | 2-4 agents across Sales and Service | $150K–$350K |
| Enterprise program | Full Agentforce transformation with Data Cloud | $400K+ |
These ranges reflect the market broadly. At Maple Wave AI, our engagements start at $65K for a production-ready single-agent deployment — we keep costs lower by running lean, specialist teams rather than large SI project structures.
Total first-year cost: a realistic example
Here is what a typical mid-market Service Cloud Agentforce deployment looks like in year one:
| Cost item | Annual cost |
|---|---|
| Agentforce licensing ($2/conversation × 5,000/mo) | $120,000 |
| Implementation (single agent, 8 weeks) | $80,000 |
| Ongoing optimization (quarterly reviews) | $24,000 |
| Total year 1 | $224,000 |
| Value delivered (at $21 fully-loaded cost per deflected case) | $1,260,000 |
| Year 1 ROI | 5.6x |
How to reduce your Agentforce costs
Start narrow, expand deliberately
The biggest cost mistake we see is companies trying to deploy a generalist agent that handles everything. A tightly scoped agent handling your top 3 use cases will deflect more volume at lower implementation cost than a broad agent that handles everything poorly.
Optimize your handoff threshold
Every conversation the agent escalates to a human is a conversation you paid $2 for and still had to resolve with human cost. Investing in reducing your escalation rate from 40% to 20% effectively doubles your ROI from the same licensing spend.
Negotiate annual volume commitments
If you can forecast your conversation volume reliably, Salesforce will typically offer discounted per-conversation rates for annual volume commitments. We have seen clients negotiate rates as low as $1.50 per conversation on committed volumes of 100,000+ conversations per year.
Bottom line
Agentforce is not cheap — but for companies with meaningful service or sales volume, the ROI math works clearly in its favor. The key is sizing the deployment correctly for your volume and not over-building in year one.
If you want a specific cost model built for your org — with your actual case volumes and agent costs plugged in — that is exactly what our free strategy call covers.