How should Salesforce price Agentforce — per agent or per seat?
Salesforce should move to a hybrid per-agent base + per-outcome model by 2027. Lock customers into $150–250/agent/month floor (predictable), then layer per-conversation overages ($0.75–1.50) for usage spikes, with per-outcome bonuses ($5–25/won-deal, case-resolved) to align margin with customer value realization.
Why $2/Conversation Won't Last
- Unpredictable CAC payback: $2/convo at 3–5 conversations/day = $150–500/agent/mo; enterprise procurement can't budget it
- Margin compression at scale: High-volume customers (1000+ agents, 50K+ conversations/month) collapse unit economics; Salesforce needs hybrid to protect blended margin
- HubSpot + Microsoft precedent: HubSpot Breeze bundled per-seat ($500–2K/mo depending on contact tier), Microsoft Copilot $30/seat/mo; per-conversation is anomaly in AI pricing
- Tenant misalignment: Customer success teams want "how much will we spend?" by budget cycle; CFOs reject variable-only models
- Seat vs. Agent semantic drift: "Agent" (active AI user per day) ≠ "seat" (licensed access); customers double-count, disputes ensue
What Salesforce Should Actually Charge
- Establish per-agent floor: $175/agent/month (paid-seat committed). 20% cheaper than per-seat licensing on raw units; locks 12-month contracts; typical org = 100–500 agents = $21K–105K ARR.
- Layer conversation overage at $1.25/convo beyond 150/agent/month. Conversation 1–150 = "included"; 151+ triggers overage. Incentivizes efficiency, protects Salesforce from 10K-convo outliers.
- Outcome commission tier: +$15 per deal-stage update escalation. Tracks when Agentforce pushes deal forward (qualification, discovery, closed); separate from conversation cost; appeals to RevOps logic ("we pay for results").
- Tiered discounts for 50K+ annual conversations. 10% off base at 50K; 15% at 250K. Enterprise scale plays; Salesforce absorbs margin hit vs. Losing to HubSpot Breeze.
- Bundled Agentforce + Data Cloud pricing. Agent needs data enrichment to work; bundle at $250/agent/month = $175 Agentforce + $75 Data Cloud. Upsells platform lock-in.
- Annual vs. Monthly cadence incentive. Monthly = $175/agent; annual pre-pay = $160/agent/month (8.6% discount). Enterprise budgeting wins; Salesforce gets cash upfront.
- Outcome-not-reached credit. If Agent generates <80 conversations/month after 90-day ramp, pro-rata credit ("we pay for utility"). Reduces churn, builds trust.
- Region-weighted pricing. US/EU = base $175; APAC/LATAM = $140/agent. Defensible by wage arbitrage; Microsoft Copilot precedent.
Pricing Model Comparison
| Pricing Model | Customer Profile | 2025 Reality | 2027 Forecast | ARR Impact (1K-agent Org) |
|---|---|---|---|---|
| Per-Conversation ($2 flat) | SMB, low-forecast use | Agentforce 2.0 baseline (launched Q1 2025) | Abandoned; too volatile | $18K–90K (highly variable) |
| Per-Seat ($30–50/mo, like Copilot) | Enterprise standard-buy | HubSpot Breeze model (bundled) | Salesforce undercuts at $35–40 | $420K/yr (predictable, low margin) |
| Hybrid: Base + Overage | Mid-market + Enterprise | Positioning gap (test now) | Salesforce standard by 2027 | $210K base + $45K–75K overage = $255K–285K |
| Per-Outcome ($5–25/action) | RevOps-native, outcome-aligned | Force Management, Pavilion POV only | Co-primary with base-plus-overage | $60K–150K (bonus pool only) |
| Per-Agent (committed seat) | Enterprise procurement | Microsoft Copilot precedent | Microsoft/Copilot Studio capture risk | $420K (defensible, competes w/ Breeze) |
Competitive Pressure & Margin Math
FAQ
Why won't Agentforce's $2-per-conversation model survive to 2027? At 3–5 conversations per agent per day, $2/convo produces $150–500/agent/month, which enterprise procurement can't budget against a predictable cycle. High-volume customers running 1,000+ agents and 50K+ conversations a month collapse Salesforce's unit economics.
The article calls it a 2025 placeholder useful for adoption ramp but unsustainable, and recommends sunsetting it by 2027.
What per-agent floor price does the article recommend? It recommends a committed per-agent floor of $175/agent/month on a paid-seat basis, which it frames as roughly 20% cheaper than per-seat licensing on raw units while locking 12-month contracts. A typical org of 100–500 agents lands at $21K–105K ARR.
Annual pre-pay drops the rate to $160/agent/month, an 8.6% discount.
When does conversation overage kick in under the hybrid proposal? Conversations 1–150 per agent per month are included, and overage of $1.25/convo triggers at 151 and beyond. This protects Salesforce from 10K-conversation outliers while incentivizing efficiency. The structure layers on top of the $175 per-agent base.
How do HubSpot Breeze and Microsoft Copilot pricing compare to Agentforce? HubSpot Breeze is bundled per-seat at roughly $500–2K/month depending on contact tier, and Microsoft Copilot is $30/seat/month. Against these, per-conversation pricing is the anomaly in AI pricing. The article wants Salesforce to undercut Breeze on per-seat dollars while mirroring Microsoft Copilot Studio's seat-plus-usage tiers.
How much ARR could the hybrid model generate per 1,000-agent org? The hybrid base-plus-overage model projects $210K base plus $45K–75K overage, totaling roughly $255K–285K per 1,000-agent org, versus $18K–90K under the volatile per-conversation model today. The bottom-line forecast pushes blended ARR to $250K–350K per 1,000-agent org by 2027–2028.
The article suggests A/B testing hybrid pricing on a mid-market cohort by Q3 2026.
Bottom Line
The $2/conversation model was a 2025 placeholder—useful for adoption ramp, but unsustainable for enterprise procurement. Salesforce owns the incumbency; it should price like it: move to per-agent committed base ($175/mo), overlay conversation overage ($1.25+) for usage spikes, and add per-outcome bonuses ($15–25) to align with RevOps margin-first mentality.
This mirrors Microsoft Copilot Studio strategy (seat + usage tiers) and undercuts HubSpot Breeze on per-seat dollar while restoring Salesforce predictability. By Q3 2026, Salesforce should A/B test hybrid pricing on mid-market cohort; sunset pure per-conversation by 2027.