In 2026, the biggest names in CRM and customer service software quietly rewrote the rules of how they charge for AI. Salesforce launched Agentforce Help Agent in July 2026 with a flat $2-per-autonomous-resolution price tag — you pay nothing if the AI can’t close the ticket on its own. HubSpot moved its Breeze Customer Agent from $1.00 per conversation to $0.50 per resolved conversation back in April. Zendesk is charging $1.50–$2.00 per automated resolution. Intercom’s Fin has been doing $0.99 per resolved conversation for a while. This isn’t a marketing gimmick — it’s a fundamental repricing of AI agents around outcomes instead of seats or usage, and it changes how every CRM buyer should evaluate, budget for, and negotiate AI tools going forward.

Key Takeaways

• Salesforce’s Agentforce Help Agent (GA July 2026) charges $2 per autonomous resolution — no charge if a human has to step in or the customer walks away unhappy.
• HubSpot cut its Breeze Customer Agent price from $1.00/conversation to $0.50 per resolved conversation, effective April 14, 2026, after resolution rates hit 65% across 8,000+ activations.
• Zendesk and Intercom’s Fin are running similar outcome-based models at roughly $1.50–$2.00 and $0.99 per resolution, respectively — though Fin is now a Salesforce subsidiary, not an independent vendor (see below).
• Salesforce now runs three concurrent Agentforce pricing models (per-conversation, Flex Credits, per-user licenses) alongside the new pay-per-resolution option — a sign the market hasn’t converged on one right answer yet.
• Outcome-based pricing removes forecasting risk for buyers but introduces a new one: vendors define “resolution,” and that definition determines your bill.
• For SMB and mid-market buyers, this is the first time AI customer service pricing maps directly to ROI instead of headcount or license count — but it requires new procurement math, not the old per-seat spreadsheet.

The $2 Question That’s Rewriting CRM Procurement

Here’s a concrete scenario finance teams are running into right now: a support team signs up for an AI agent, gets billed $2 for every ticket it closes on its own, and pays literally nothing in months where volume is low. Compare that to the alternative most of these same vendors were selling eighteen months ago — a flat monthly fee per seat, or a metered credit system where costs could swing wildly and unpredictably based on token consumption. That shift, from “pay for access” to “pay for results,” is exactly what happened across the CRM AI landscape between late 2024 and mid-2026, and it’s arguably a bigger deal for buyers than any individual feature announcement.

Salesforce is the clearest example. When Agentforce launched at Dreamforce 2024, it was priced at $2 per conversation — meaning you paid whether or not the AI actually solved the customer’s problem. That model drew real criticism. Capgemini’s Timo Kovala described it as limiting flexibility and obscuring costs, and Elements.cloud CEO Ian Gotts said customers wanted “caps and predictable ROI” instead of what amounted to a blank check. Salesforce responded in May 2025 with consumption-based Flex Credits, then in late 2025 added traditional per-user licenses. Now, with Agentforce Help Agent generally available as of July 2026, there’s a fourth option: pure pay-per-resolution, where the meter only runs when the AI actually finishes the job — handling a case, rescheduling an appointment, updating an order — start to finish, with no human intervention and no unhappy customer escalation.

HubSpot made a nearly identical move on its own timeline. Its Breeze Customer Agent was billed at $1.00 per conversation; as of April 14, 2026, it dropped to $0.50 per resolved conversation. HubSpot’s chief customer officer, Jon Dick, put the logic plainly: “Outcome-based pricing removes that risk. You pay when it works, full stop.” The company backed the change with real performance data — Breeze was resolving 65% of conversations and had cut resolution time by 39% across more than 8,000 activations, giving it enough confidence in the product to bet its revenue on outcomes rather than volume.

Zendesk and Intercom are playing the same game with slightly different numbers: Zendesk charges $1.50 per automated resolution on committed volume (or $2.00 pay-as-you-go), while Intercom’s Fin AI Agent has held at $0.99 per resolved conversation. Worth noting: Salesforce acquired Intercom’s Fin business for roughly $3.6 billion in June 2026, so one of the four vendors in this pricing comparison is now also a Salesforce subsidiary — a detail that matters for anyone evaluating these as genuinely independent competitors. Sierra AI, a newer entrant built entirely around outcome-based pricing, has reportedly grown past $150M in ARR without ever selling a seat.

Why This Matters for CRM and ERP Buyers

For years, CRM software pricing was straightforward, if not always cheap: you paid per seat, per month, and your costs scaled with headcount. AI agents broke that model in an uncomfortable way — a single AI “agent” can do the work of several human reps, so seat-based pricing either undercharges vendors dramatically or forces buyers into confusing consumption-based credit systems where nobody can predict next month’s invoice.

Outcome-based pricing solves a real problem on both sides. Vendors get paid in proportion to the value they actually deliver, which pushes them to keep improving resolution rates instead of just shipping more features. Buyers get a cost structure that’s directly traceable to business results — a finance team can say “we spent $4,000 last month and resolved 2,000 tickets” instead of trying to justify a seat count against usage that doesn’t map cleanly to work performed.

But the model isn’t free of risk. The entire cost structure hinges on how the vendor defines “resolution” — Zendesk and others often use a quiet-period definition (e.g., the customer doesn’t reopen the ticket within 72 hours), which is reasonable but means the vendor, not you, controls the metric your bill depends on. Analysts have also flagged a genuine adversarial risk: a bad actor (or a customer gaming a refund policy) could deliberately manufacture negative interactions to inflate a company’s AI costs, since pricing and behavior are now directly linked in a way seat licenses never were.

Practical Use Cases Across Industries

• Distribution and manufacturing: A distributor fielding order-status and return questions can deploy an AI agent priced per resolution, paying only for the volume of tickets it actually closes — useful for businesses with seasonal support spikes where a flat license fee would go unused most of the year.
• Professional services and healthcare admin: Appointment rescheduling and intake questions are exactly the kind of repeatable, well-defined workflow outcome-based pricing is built for, per Zendesk’s own guidance that this model works best for “well-defined, repeatable workflows.”
• Nonprofits and education: Organizations with tight, unpredictable budgets can adopt AI support without committing to a large seat-based contract, paying incrementally as the AI proves it can resolve donor or student inquiries.
• Sales and marketing: HubSpot’s Prospecting Agent shifted to $1 per recommended lead instead of a monthly per-contact fee — a model that rewards actual pipeline contribution over enrollment volume.

Benefits and Challenges

Benefits:

• Costs scale with value delivered, not headcount or raw activity
• Removes the “will this AI agent even get used” risk that comes with flat licensing
• Forces vendors to compete on resolution quality, not just feature lists
• Easier to model ROI for a CFO than seat-based AI add-ons

Challenges:

• The vendor defines “resolution” — read the fine print on escalation and reopen windows
• Harder to forecast costs during unpredictable volume spikes (a viral complaint, a product recall)
• Multiple pricing models across the same vendor’s product line (Salesforce now runs four) can make apples-to-apples comparison genuinely difficult
• Requires clean underlying data and well-scoped workflows; outcome pricing on ambiguous tasks (strategic advice, open-ended research) doesn’t have a clean unit to charge for

Implementation Best Practices — and Common Mistakes

Best practices:

1. Get the vendor’s exact definition of “resolution” in writing before signing — including reopen windows and what counts as escalation.
2. Model your expected ticket/conversation volume against both the old and new pricing structures before switching; outcome pricing isn’t automatically cheaper.
3. Start with a narrow, well-defined workflow (order status, appointment scheduling, password resets) rather than open-ended support, since that’s where outcome pricing performs predictably.
4. Set internal alerting on resolution-driven spend the same way you’d monitor cloud compute costs — it’s consumption-based even when it looks flat-rate.

Common mistakes:

• Assuming “pay per resolution” means “cheap by default” — a high-resolution-rate agent handling high ticket volume can cost more than a seat-based plan.
• Deploying the AI agent across every channel on day one instead of piloting on one workflow to validate the vendor’s resolution definition against your own definition of “solved.”
• Ignoring the escalation path economics — if your team has to clean up after low-quality “resolutions,” the per-ticket savings evaporate into rework.

Comparison: Outcome-Based AI Pricing Across Major CRM Vendors

CRM Experts Online’s Perspective

We implement these platforms for clients, not just read about them, and this pricing shift changes real conversations we’re having in scoping calls this quarter. The instinct for a lot of business owners is to treat “pay per resolution” as an automatic cost win compared to seat licenses — it isn’t automatically either way. We’ve started running side-by-side cost models for clients evaluating Agentforce Help Agent against their existing Service Cloud seat spend, and the honest answer depends entirely on ticket volume, complexity, and how aggressively the AI can resolve without human handoff in your specific workflows.

The bigger opportunity we see is for mid-market companies that were priced out of enterprise AI service tools because seat-based AI add-ons didn’t make sense against their support volume. A distributor or professional services firm with seasonal ticket spikes can now pilot Agentforce Help Agent or Breeze Customer Agent without committing to a flat annual license — and if it doesn’t perform, the downside is bounded by what it actually resolved, not a sunk license cost. That’s a genuinely different conversation than we were having a year ago, and it’s worth revisiting your AI service roadmap even if you evaluated these tools before and passed.

FAQ

Is pay-per-resolution pricing actually cheaper than seat-based licensing? Not automatically. It depends on your ticket volume and the AI’s real resolution rate for your specific workflows. High-volume, high-resolution-rate deployments can cost more per month than a flat seat license would have.

Who decides what counts as a “resolution”? The vendor does, contractually. Most use a quiet-period definition (e.g., the customer doesn’t reopen the ticket within a set window), but the exact terms vary by vendor and should be reviewed before signing.

What happens if the AI agent fails to resolve an issue? With these models, you’re not charged. Salesforce, HubSpot, Zendesk, and Intercom all structure pricing so escalations to a human or unresolved tickets don’t generate a fee.

Can I mix pricing models within the same vendor? In Salesforce’s case, yes — the company currently offers per-conversation, Flex Credits, per-user licenses, and pay-per-resolution simultaneously, and customers can choose based on their use case.

Is this model available for sales or marketing AI agents, not just support? Partially. HubSpot’s Prospecting Agent has moved to a similar outcome-based structure ($1 per recommended lead), but outcome pricing works best for well-defined, repeatable tasks — it’s a poorer fit for open-ended work like strategic research or content generation.

Does outcome-based pricing create any new risks? Yes — cost predictability during volume spikes, and a theoretical exposure to bad actors manipulating outcomes (e.g., manufacturing dissatisfaction) to drive up automated resolution counts or force escalations.

Should smaller businesses consider this before larger enterprises? If anything, it’s a better entry point for smaller businesses — it removes the large upfront license commitment that made enterprise AI service tools inaccessible to smaller support teams.

How do I know if my support workflows are a good fit for outcome-based pricing? Look for high-volume, well-defined, repeatable requests — order status, appointment scheduling, account questions. Ambiguous or judgment-heavy workflows don’t have a clean “resolved” line and are harder to price or evaluate this way.

Conclusion

The move to pay-per-resolution pricing across Salesforce, HubSpot, Zendesk, and Intercom isn’t a side detail in the AI-agent story — it’s a signal that these vendors are confident enough in their AI’s actual performance to bet their revenue on it, and it opens a real opportunity for businesses that couldn’t justify seat-based AI licensing before. The catch is that evaluating these tools now requires modeling ticket volume and resolution rates, not just comparing sticker prices. If you’re weighing Agentforce Help Agent, Breeze Customer Agent, or a similar tool against your current CRM setup, CRM Experts Online can run the real numbers against your ticket volume and workflows and help you pilot the right one — schedule a consultation with our team to get a cost model built around your actual support data before you sign anything.