So HubSpot just made a big pricing change to its Breeze AI agents.
Starting April 14, its AI Customer Agent pricing moves from $1.00 per conversation to $0.50 per resolved conversation. And its Prospecting Agent moves from a flat monthly charge per contact enrolled to $1.00 per lead recommended for outreach.
On its face, this seems like a big deal, especially the first change — moving to outcome-based pricing. You only pay when the AI actually does its job. No more paying for failed attempts, dead-end conversations, or leads that go nowhere. As HubSpot’s Chief Customer Officer Jon Dick noted: “You pay when it works, full stop.”
The numbers back up the change. HubSpot says Breeze Customer Agent already resolves 65% of conversations across 8,000+ customers, and cuts resolution time by 39%. Prospecting Agent activations are up 57% quarter-over-quarter. These are good numbers and certainly radically higher than when the agentic products launched.
But here’s the thing. Is this actually going to matter in 12 months?
So much of pricing is just a game in the end. And pricing that looks like product innovation often when you dig deeper, is a bit less innovative for the customers than in looks.
Sierra Was Way Early Here. And It Worked.
HubSpot isn’t pioneering outcome-based pricing for AI agents. Sierra arguably did, albeit in the enterprise.
Bret Taylor and Clay Bavor launched Sierra in early 2024 with outcome-based pricing baked in from day one. Their model: you pay a pre-negotiated rate when the AI agent resolves an issue autonomously. If it has to escalate to a human, it’s free. As Taylor put it at Sequoia’s AI Ascent: “If you’re selling software that completes a job, what is the secular business model for that? Let’s pay for a job well done.”
And the results have been extremely successful.
Sierra hit $100M ARR in just 21 months. They then posted its first $50M quarter, crossing $150M+ ARR by February 2026. $10 billion valuation. One in four customers has revenue over $10 billion. Customers like ADT, SiriusXM, Rivian, and Cigna seeing 50-90% of customer service interactions fully automated. They acquired Opera Tech in Tokyo in March and are expanding offices globally.
Taylor has been vocal about why this matters. As recently as his March 2026 conversation with John Collison on Cheeky Pint, he laid it out clearly: the whole market is going to go towards agents and towards outcomes-based pricing. In his words, “it’s just so obviously the correct way to build and sell software.” Taylor was both CTO of Facebook and co-CEO of Salesforce.
And agrumbly, outcome-based pricing gives startups a structural advantage over incumbents. Legacy CX providers face a dilemma because their revenue models depend primarily on seat-based pricing. The more effective their AI becomes, the fewer seats their clients need. That undermines the provider’s own revenue. It’s a real issue and anyone that is invested in legacy CX players sees seat decay as a massive issue.
Sierra, with no reliance on seat-based pricing, had no conflicting incentives. The AI works better, Sierra makes more money.
Intercom Did It, Too. And Grew Fin From $1M to $100M+ ARR on It.
Intercom rebooted its entirely company around its Fin AI agent and launched with $0.99 per resolution pricing. It has been one of the clearest success stories in B2B for the model. As of April 2026, roughly 8,000 companies use Fin. It now resolves 2 million customer issues per week (up from 1 million just months ago) and grew from $1M to $100M+ ARR. Intercom’s monthly growth rate swung from as low as 4% to over 37%, and the company shifted 80% of its R&D dollars to Fin and grew its AI team from 6 to 60 people in three years.
The pricing is straightforward: $0.99 per resolution, on top of seat-based plans (Essential at $29/seat/month, Advanced at $85, Expert at $132). You only get charged when Fin actually resolves the conversation. If the customer confirms the answer helped, or doesn’t request more help after Fin answers, that’s a billable resolution. If a human agent steps in before resolution, no charge. One charge per conversation, even if Fin handles multiple questions in the same thread.
Intercom’s backs it with a $1M performance guarantee. If you sign up for their Fin Guarantee Success Program and don’t achieve at least a 65% resolution rate, they’ll pay you $1M. As Intercom’s president Archana Agrawal described it, outcome-based pricing exposed every weak link in their organization. Sales couldn’t optimize for licenses. CS couldn’t hide behind usage. The product had to actually work, consistently. It was a forcing function.
And Intercom launched the Fin API Platform, opening Fin up to third-party developers for the first time. It’s available to customers spending at least $250K on Intercom. They’ve built their own in-house LLM called Fin Apex, which they claim outperforms GPT-5.4 and Claude Sonnet 4.6 on resolution rate and hallucinations. CEO Eoghan McCabe isn’t being subtle about the stakes: “If you can’t become an agent company, your CRUD app business has a diminishing future.”
Fin’s resolution rates tell the story of the broader AI improvement trajectory. Fin launched at roughly 27% resolution rates. The average across Intercom’s customer base is now 66-67%. That climb from 27% to 67% in under two years is the whole game. And also … why pricing per resolution in the end may not be as disruptive as it looks at first.
Salesforce Went a Different Way. Three Pricing Models at Once.
Meanwhile, Salesforce has been on its own wild ride with Agentforce pricing. We wrote about this recently. They shipped three different pricing models in roughly 18 months:
October 2024: $2 per conversation. Elegant on a whiteboard, tougher in production. What even counts as a “conversation” when a single customer query triggers 8 backend processes? The original $2/conversation model also priced out SMBs and nonprofits. Only about 3,000 of 5,000 initial Agentforce deals were paid.
May 2025: Flex Credits at $0.10 per action. More granular. More aligned with actual work done. But still fundamentally consumption-based, still unpredictable for enterprise procurement.
Late 2025: Per-user licenses at $125+/month. Seats become the primary wrapper again. CFOs get a number they can budget. CIOs get a contract shape they understand.
All three run simultaneously now. Many people look at this and see complexity. But I’d argue it’s still smart and responsive to where customers are, right now, in their agentic journey. For now. When the market hasn’t converged on how to buy something, you don’t force one model. You let customers self-select.
The result? Agentforce hit $540M ARR by Q3 FY2026, growing 330% year-over-year. 18,500 total deals, 9,500 paid. Only about 8% of Salesforce’s 150,000+ customer base has adopted so far. Having three on-ramps instead of one turned out to be the right call.
Salesforce also arguably has to run three pricing models because they can’t fully commit to outcomes-based pricing without cannibalizing their own seat-based revenue. Sierra doesn’t have that problem. Intercom is navigating it. And now HubSpot is joining that conversation.
Zendesk Jumped In Too
Zendesk has gone all-in on outcome-based pricing through what they call “Automated Resolutions.” The model launched at their 2025 Relate Conference as part of the Zendesk Resolution Platform, and as of April 2026 it works like this: every Suite and Support plan includes a baseline number of free automated resolutions per agent per month (varying by tier, up to 10,000 allocated per year). Beyond that, you pay $1.50 per automated resolution on committed volume, or $2.00 per resolution on pay-as-you-go. An automated resolution only counts when the AI fully resolves a customer’s issue without human intervention, confirmed after 72 hours of inactivity.
On top of that, the real AI muscle requires the Advanced AI add-on at $50/agent/month, which gets you intelligent triage, generative replies, and the agent copilot. Suite plans themselves range from $55/agent/month (Team) to $169/agent/month (Enterprise). So the total stack for a team running serious AI automation is: seat license + Advanced AI add-on + per-resolution fees. A 20-agent team on Suite Professional with active AI easily runs $75,000-$100,000/year.
Zendesk also introduced an “AI Dynamic Pricing Plan” for enterprises that lets you shift committed budget between human agent seats and AI-powered automated resolutions as your strategy evolves, without renegotiating contracts. It’s the closest thing in the market to acknowledging the seat cannibalization problem head-on.
One Zendesk customer reportedly burned through a year’s worth of automated resolutions in just a few weeks. And Trustpilot reviews aren’t kind about the layered cost structure: seat fees plus AI add-on fees plus per-resolution fees. As one customer put it, providing only 15 resolution credits as a baseline and then charging on top of a fixed license cost plus a fixed AI add-on cost is “too much.” That’s the double-edged sword of outcome-based pricing layered on top of legacy seat pricing rather than replacing it.
So Here’s the Pricing Landscape as of April 2026
HubSpot at $0.50/resolution is genuinely the cheapest option in the market right now. That’s a real competitive wedge for SMBs.
But The Math Changes Fast When Resolution Rates Hit 90%
Right now, AI resolution rates in customer support are all over the map. HubSpot is at 65%. Intercom’s Fin averages 66-67% across its 8,000 customers. Sierra’s best deployments hit 90%. Many early AI agents in 2024-2025 were stuck in the low 20s. At those early rates, per-resolution pricing was a genuine gift to customers, or at least, a confidence-builder when the models were less reliable. You were essentially saying: “We know this thing fails a lot. We’ll eat the cost of the failures.”
But AI is getting better fast. Gartner projects AI agents will autonomously resolve 80% of common customer service issues by 2029. But it probably will be faster. Look at the trajectory — and the rate at which LLMs are improving. Top performers are already pushing toward 90% today. Intercom’s own Fin went from 27% to 67% in under two years. Resolution rates grew 52% in just the year between Fin 1 and Fin 2. That trajectory doesn’t stop at 67%.
When resolution rates were 25%, paying per resolution meant you paid for 1 in 4 attempts. That’s a real discount. When resolution rates hit 90%, you’re paying for 9 out of 10 attempts. At that point, per-resolution pricing and per-conversation pricing are almost the same thing.
The delta between “per use” and “per resolution” shrinks to near zero as models improve. It starts to become form over substance.
The Variable Bill Problem Hasn’t Gone Away Either
Variable bills are also a headache for CFOs and budgets. They’re aren’t such a gift in many cases. Seats do have the benefit of being predictable.
Per-resolution pricing sounds customer-friendly until the bills show up. Intercom customers running 5,000 AI-resolved tickets monthly pay $4,950 in AI costs on top of base seat pricing. Bills can fluctuate dramatically. One month $300, the next $800, especially during high-volume periods. Community discussions are full of complaints about unpredictable costs. Some Intercom customers are already negotiating for resolution caps or flat-rate tiers because the per-resolution costs get too unpredictable at scale. And Zendesk’s triple-layer structure (seat + AI add-on + per-resolution fees) compounds the problem. A team automating 3,000 resolutions per month on Zendesk can add $3,000-$6,000/month on top of everything else.
HubSpot’s rate nominally is the best in market at $0.50. But do the math on a real deployment. 5,000 support conversations a month. At 65% resolution rate and $0.50 per resolution, that’s $1,625/month. Push that to 80%? $2,000/month. At 90%? $2,250/month. For a company growing fast, that number keeps climbing.
This is the same tension we saw with usage-based pricing in B2B broadly. Snowflake, Twilio, the whole consumption-based wave. Wall Street loved it on the way up, punished it on the way down. Customers liked the idea of paying for what they use until they realized they couldn’t predict what they’d use.
Per-resolution AI pricing has the same DNA. It’s aligned. Until it’s not.
HubSpot Is Still Smart to Do This Now
All that said, I think HubSpot is making the right move. Here’s why:
1. It’s a great wedge for new buyers. At $0.50/resolution vs. Intercom at $0.99, Zendesk at $1.50-$2.00 plus a $50/agent AI add-on, and Salesforce’s zigzag, HubSpot is the cheapest option and it’s baked into the platform rather than stacked as an add-on. For an SMB making their first AI bet, that matters a lot. Zendesk’s triple-layer cost structure in particular makes HubSpot look elegant by comparison.
2. It reduces the “trust gap” for AI skeptics. A lot of potential AI buyers are still scared and/or skeptical of AI Agents. They’ve been burned by chatbots that don’t work. Per-resolution pricing gives them a safety net or sorts. If it doesn’t work, you don’t pay. HubSpot is even throwing in a free 28-day trial. For SMBs who haven’t tried AI agents yet, the trial is probably more important than the pricing model itself. Get them in the door. Let them see 65% resolution rates. The pricing is secondary to the habit formation.
3. Sierra proved the model. Now it’s table stakes. Sierra showed that outcome-based pricing can power one of the fastest-growing enterprise software companies ever. Intercom showed it can work layered on top of an existing seat model. HubSpot is smart to follow. But they’re following, not leading. This would have been a much bolder move 18 months ago.
4. HubSpot has a data moat argument. Their pitch is that Breeze agents work better because they’re built into the CRM with full customer context, relationship history, the whole deal. Outcome-based pricing is a way of saying “we’re so confident in our context advantage that we’ll bet our revenue on it.” Smart positioning, even if it’s the same argument Sierra, Intercom, and everyone else is also making.
This Would Have Been a Much Bigger Deal in 2025. It May Not Even Matter in 2027.
If HubSpot had made this move 12 months ago, when AI resolution rates were 20-30% and everyone was still figuring out if these things even worked, per-resolution pricing would have been a seismic shift. A pure bet-on-ourselves move when nobody else was willing to.
In April 2026, with Sierra at $150M+ ARR on pure outcome pricing, Intercom at $100M+ ARR on per-resolution, Salesforce running three models simultaneously, and Zendesk already having switched, the move is more expected than bold. Resolution rates are climbing past 65% and heading to 80-90%. The risk HubSpot is taking is smaller because the technology has caught up.
And asTaylor noted: as AI agents get better and better, almost every resolved conversation is a conversation that would have been a successful interaction anyway. The distinction between “per use” and “per resolution” disappears.
The real question isn’t per-use vs. per-resolution anymore. The real question is what happens when resolution rates are so high that the AI is handling almost everything.
Do you go back to flat-rate pricing? Do you charge per seat because the “agent” is now the AI, not the human? Do you just raise the platform subscription and bundle it all in?
Taylor says outcome-based pricing is the end state. He may be right in the long run. But in the medium term, Salesforce’s messy-but-pragmatic three-model approach might be the most honest response to a market that hasn’t figured out how it wants to buy yet.
My bet: in 2-3 years, per-resolution pricing will be a footnote for most categories. The models will be so good that charging per resolution will be like charging per successful Google search. The success rate will be so high that the distinction between “attempt” and “resolution” won’t carry economic weight.
