Agentforce wins for Salesforce-locked enterprises. The per-conversation pricing aligns with deflection ROI. After 90 days of real use, here's the honest breakdown.
What Agentforce is in 2026
Agentforce is Salesforce's platform for building and deploying AI agents inside your Salesforce instance. Three flavors:
- Agentforce Service — CS deflection agents
- Agentforce Sales — sales reps and prospecting helpers
- Custom Agents — built with Agentforce Studio for specific workflows
| Agentforce | |
|---|---|
| Maker | Salesforce |
| Pricing | $2 per conversation (base) |
| Best for | Salesforce customers |
| Worst at | Non-Salesforce environments |
| Agent Rank | A-tier (76/100) |
90 days of real use — the breakdown
We deployed Agentforce Service for a mid-market SaaS company (~$50M ARR) over 90 days. Results:
| Metric | Result |
|---|---|
| Deflection rate (tier-1) | 62% |
| Customer satisfaction (CSAT) | 4.2/5 |
| Avg resolution time | 90 seconds (vs human 4 min) |
| Cost per resolution | $2 (Agentforce) vs $8 (human tier-1) |
| Issues escalated to humans | 38% (mostly nuanced) |
What Agentforce does well
Native Salesforce data grounding. Agents have instant access to customer records, opportunity history, and support tickets without integration plumbing.
Outcome-based pricing. You pay per conversation resolved. If the agent fails, you don't pay. Aligns vendor incentives with customer outcomes.
Compliance inheritance. Inherits Salesforce SOC 2, audit logs, encryption. Easier enterprise security review.
Multi-agent orchestration. Service + Sales + Custom agents can hand off naturally within Salesforce.
Where Agentforce falls short
Outside Salesforce, value drops sharply. If your support workflows live in Zendesk-only or Intercom-only, Agentforce isn't the right tool.
Per-conversation pricing balloons at scale. At 50K+ conversations/month, per-seat tools (Decagon, Sierra) often work out cheaper.
Agent quality varies by use case. Service agents are mature; Sales agents and Custom agents are newer with more rough edges.
Setup is real work. "Native" doesn't mean "no setup" — expect 4-8 weeks of agent training and tuning.
Three patterns from real deployments
Pattern 1: Service first, expand to Sales. Most Agentforce wins start with Service deflection. Sales agents come later.
Pattern 2: Hybrid with Sierra. Some enterprises run Agentforce for Salesforce-grounded queries and Sierra for brand-led complex CS.
Pattern 3: Don't customize before you measure. Start with default Service agent; measure baseline; customize only when specific failures emerge.
The verdict
Worth it if:
- You're a Salesforce customer (table stakes)
- You have measurable deflection-rate goals
- You can commit 4-8 weeks to setup
- Outcome-based pricing model fits your finance team
Skip if:
- You're not on Salesforce
- Your support volume is under 5K conversations/month
- You need a customer-facing brand-voice agent (use Sierra)
For comparisons see Sierra, Decagon, and The 15 best AI agents of 2026.