On Monday, HubSpot flipped the switch on outcome-based pricing for its Breeze AI agents. The Customer Agent now costs $0.50 per resolved conversation — down from $1.00 per conversation, resolved or not. The Prospecting Agent moved from a flat monthly fee per enrolled contact to $1.00 per lead recommended for outreach. “You pay when it works, full stop,” said Jon Dick, HubSpot’s Chief Customer Officer.
That’s half of what Intercom charges for Fin ($0.99 per resolution). It’s a third of Zendesk’s committed rate ($1.50 per automated resolution) and a quarter of its pay-as-you-go price ($2.00). Add Sierra, which pioneered outcome pricing in the enterprise — custom rates, contracts starting around $150K/year — and you have something that didn’t exist six months ago: a competitive market with published prices for the same unit of work.
We’re watching a commodity market form in real time.
Four vendors, one unit
Consider what these companies are actually selling. Not software access. Not seats. Not platform time. A resolved customer conversation — a discrete, measurable outcome that used to require a human earning $35K-$55K per year.
The pricing spectrum as of this week:
- HubSpot Breeze: $0.50/resolution (65% resolution rate across 8,000+ customers)
- Intercom Fin: $0.99/resolution (2 million resolutions/week, up from 1 million months ago, $100M+ ARR)
- Zendesk: $1.50/resolution committed, $2.00 pay-as-you-go (bundled with per-agent suite plans from $55-$169/month)
- Sierra: Custom enterprise rates ($150M+ ARR, $10B valuation, clients including ADT, SiriusXM, Rivian)
That’s a 4x spread between cheapest and most expensive. In traditional SaaS, per-seat pricing varied by maybe 2x across competitors for similar functionality. Outcome pricing compresses the range faster because the buyer can directly compare: one resolution at Vendor A costs $X, at Vendor B it costs $Y. No implementation complexity or feature matrices to hide behind.
The deflationary logic
HubSpot’s move reveals a structural dynamic. When you price on outcomes, you invite direct comparison on a standardized unit. Buyers don’t need to decode what a “seat” includes or whether “Professional” covers the features they need. They ask one question: what does a resolved conversation cost?
That question creates downward pressure. HubSpot could undercut Intercom because its agents are embedded in its CRM — the context is already there, reducing inference costs per resolution. Intercom’s edge is resolution volume and rate: 2 million per week, growing fast. Zendesk charges more but bundles its per-agent platform fees, effectively cross-subsidizing from legacy seat revenue.
Each vendor’s cost structure is different, but the output is identical. That’s the textbook definition of a commodity.
The implication for the hypothesis is direct. Post 5 on this blog asked why outcome pricing was emerging — the answer was COGS math. AI inference costs made fixed-fee models untenable when the vendor bears variable compute costs per interaction. This is what happens next: once the pricing model exists, competition drives the per-unit price toward marginal cost.
But the margins tell a more complicated story
Sierra hit $100M ARR in 21 months and crossed $150M+ by February 2026. Intercom grew Fin from $1M to $100M+ ARR and now handles 2 million resolutions weekly. These aren’t collapsing businesses. They’re growing faster than most traditional SaaS companies ever did.
The reason: outcome pricing sells into a different budget. A customer support team of ten people costs $400K-$550K per year in loaded compensation. An AI agent handling 65-80% of their volume at $0.50-$0.99 per resolution might cost $50K-$150K — depending on volume — while the remaining staff handle escalations and complex cases. The buyer compares the AI cost against payroll, not against the old SaaS subscription. The addressable market expands because the reference price is human labor, not software.
Sierra’s client base — one in four customers has revenue over $10B — suggests this math works at enterprise scale. Bret Taylor told Skift last week that “the entire market is going towards agents and towards outcome-based pricing.” That’s not a prediction from a startup founder; that’s an observation from the OpenAI board chair running a company with $150M+ ARR built entirely on the model.
What the price curve means for the hypothesis
The hypothesis says software loses value as AI commoditizes production. Here’s where the evidence gets uncomfortable for clean narratives.
Per-resolution prices are falling. HubSpot’s $0.50 would have been unthinkable a year ago. That’s verification: the unit of software value is compressing, exactly as predicted.
But total vendor revenue is rising, because volume scales with price compression. Intercom doubled its weekly resolutions in months. The cheaper it gets, the more conversations companies route to AI instead of humans. Lower price per unit, vastly more units.
This mirrors a pattern from another industry that went through commodity pricing: cloud compute. AWS made server time cheap. The price per compute-hour fell relentlessly. AWS’s revenue grew relentlessly too, because the volume of compute purchased exploded. The commodity became the foundation for a larger market.
The question is whether AI resolutions follow the same curve or whether they hit a ceiling. Cloud compute scales with application growth — more apps, more compute. AI resolutions scale with… the number of customer conversations a company has. That has a natural ceiling tied to customer base size. Once you’re resolving 80% of conversations automatically, growth comes only from acquiring more customers or expanding to new use cases (sales, onboarding, retention).
Hypothesis update
The evidence from this week partially verifies and partially complicates the hypothesis.
Verified: The unit price of a software-delivered outcome is compressing. Four vendors now compete on a standardized unit, and the cheapest is half the price of six months ago. Commodity dynamics are present and accelerating.
Complicated: Total addressable market for outcome-priced software may be larger than for seat-based software, because the reference price shifts from software budgets to labor budgets. The same deflationary force that compresses per-unit price could expand total vendor revenue.
Watch next: Q2 earnings from HubSpot (May) and the next Intercom update will reveal whether HubSpot’s price cut accelerates adoption enough to grow revenue, or whether it just transfers margin to buyers. If HubSpot’s AI agent revenue grows despite the 50% price cut, the “cloud compute” analogy holds. If it stalls, per-resolution pricing is a race to the bottom with no escape.