Log entry Nº 678 · the stack watch
Bearing W 270° · November 18, 2026 · 4 min read
AI in MarketingW 270°Watch log
Agentforce Pricing Shifts Again: What Outcome-Based AI Deals Hide
Only 22.1% of IT leaders want consumption-based AI pricing, yet Salesforce's Agentforce and Horizon Media's Blu show vendors metering anyway — and defining 'done' themselves.
By James Calloway4 min read729 words

Only 22.1% of IT leaders prefer consumption-based pricing for AI and software, yet that's exactly the model SaaS vendors keep pushing as they bolt agentic AI onto their platforms. The number comes from Infotech's Global IT Leadership Survey, cited in a new analysis by Brian Jackson, principal research director at Infotech, and it frames an uncomfortable truth: the AI pricing conversation is being driven by sellers, not buyers.
Jackson's argument starts, oddly, with balloons. A store couldn't inflate his party balloons because its industrial helium tank ran empty; the workaround was buying a smaller tank and doing it himself. He left, ordered pre-inflated balloons online, and paid more for zero effort and a guaranteed outcome. Three procurement models, one birthday party — and, he argues, a clean map of agentic AI purchasing.
What are the three agentic AI pricing models?
Jackson lays out the trade-offs in balloon terms that translate directly to inference compute:
- Flat-rate pricing. Cheapest and predictable, but you do the work — and you might run out of capacity before the job is done.
- Consumption-based pricing. Metered until the task completes; budget less predictable, and you might pay for wasted runs.
- Outcome-based pricing. Most expensive, least work, outcome guaranteed. But defining "the outcome" is where the fights begin.
A vendor's heritage shapes what it offers, Jackson notes. SaaS vendors layer consumption metering onto subscriptions; business process outsourcers push outcome-based deals. That gives buyers a choice — and, more importantly, a leverage point.
How does Salesforce price Agentforce now?
Salesforce, the vendor that popularized cloud CRM, has changed Agentforce pricing multiple times in recent months as it navigates toward AI-driven CRM. The current menu has three flavors:
- Per conversation, the oldest approach, launched with Agentforce in 2024 — and only for customer-facing interactions.
- Price per resolution, available only for Help Agent, the pre-packaged agent for resolving customer inquiries.
- Flex credits, which cover both customer-facing and internal agents, with a standard credit cost per action.
Salesforce calls price-per-resolution "outcome-based pricing," and there is a genuine risk shift here: the company won't charge when a customer escalates to a human agent. But read the fine print. Salesforce defines what counts as a resolution and imposes caps — including a 10-minute limit on voice interaction before the meter ticks up. The vendor sets the definition of "done." Remember that.
Negotiation targets Jackson identifies: volume discounts on flex credits committed in advance (more credits, longer contracts, steeper discounts), whether to pay the per-seat add-on for unmetered use, and whether unused credits roll over.
What happens when clients dispute the meter?
Horizon Media offers a cautionary tale from the demand side. The agency launched agentic buying capabilities on its Blu platform in July, packaging agentic AI, data sources and analytics so clients build their own audience groups and deliver ads.
Because clients run their own analysis, they hit variable costs — and they quibble. The result, per one executive, is constant negotiation, as many as "dozens of times a week with each client." Clients argue they shouldn't pay when they perceive no value; the agency, which already ate the inference and data-call costs, counters that users need more training. Most disputes settle with the client paying most, but not all, of the consumption costs.
That's outcome-based pricing in the wild: a standing argument over what was delivered.
Where does buyer leverage actually sit?
For balloons, verification was trivial — it floats or it doesn't. Agentic AI isn't so clean-cut, Jackson writes. The dispute space in the contract is precisely where a buyer's leverage lives: the definition of an outcome, or what exactly gets metered. Focus negotiation efforts there and push back.
Nobody examined his balloon invoice at the party, he concedes. Enterprise buyers won't get that luxury. With agentic AI at the core of operations, the CFO will review the bill, and the vendor's definition of "done" will matter as much as the price. Know your leverage points before signing.
The survey backdrop sharpens the point: subscription per user still leads at 36.4%, open source with support contracts follows at 23.1%, consumption pricing trails at 22.1%, and one-time purchase sits at 13.9%. Buyers still prefer predictability. Vendors keep selling meters.
Plot points
- Infotech's Global IT Leadership Survey: subscription per user leads at 36.4%, open source plus support at 23.1%, consumption pricing at 22.1%, one-time purchase at 13.9%.
- Salesforce's price-per-resolution model for Agentforce caps voice interactions at 10 minutes before the meter ticks up.
- Horizon Media's Blu agentic buying platform launched in July and triggers negotiations 'dozens of times a week with each client.'
- Agentforce pricing has changed multiple times since the per-conversation model launched in 2024.
via techtarget.com (Source)
More from James Calloway
Show full bio
Correspondent covering people, launches and day-to-day practice in marketing technology at Martech Beacon.
33 articles