Content oversight provided by Studio 1847
It is not just major software companies building AI agents today; the market ranges from early-stage startups to legacy industry giants. Contract values run from a few dollars to six-figure enterprise deals as usage scales. Vendors are exploring various pricing models, including outcome-based approaches, but no industry standard has emerged. As organizations pour millions into AI for customer support, sales, and operations, most are left guessing whether they are actually getting their money’s worth.
“Pricing is already hard,” says Deniz Akbasaran, a leading pricing and monetization strategist who shapes commercial frameworks for AI products at Gorgias, a B2B software ecosystem serving over 16,000 global e-commerce brands. “Pricing AI agents is harder because there isn’t yet a proven playbook. Industry leaders are essentially figuring it out as they go.”
More Top Picks Best Memory Foam Wheelchair Cushions For Elderly
With organizations investing in AI racing to monetize it, strategists are building the financial rules for the sector from scratch.
The resolution paradox
A fundamental paradox underlies the current AI boom, though vendors seldom address it openly. The prevailing model for AI customer support agents is resolution-based billing, where companies pay only when the AI independently resolves an issue. While logical on paper, Akbasaran has identified a major operational flaw.
“Improving agent behavior sometimes means escalating more complex cases to human agents,” Akbasaran explains. “But when you bill on a resolution-based model, every escalation directly impacts your revenue. You end up with a fundamental tradeoff between quality and monetization that most software providers haven’t figured out how to resolve.”
This is why some vendors switched to per-conversation pricing: Every conversation incurs an LLM cost, so leaving it unbilled hurts margins and sidesteps the whole debate over what counts as a “successful resolution.” The catch is that you then have to prove every single conversation adds value, even the ones handed over to a human, and that is a much harder sell to corporate buyers.
Unpredictable margins and data volatility
Traditional software vendors know their cost per seat. AI agent vendors don’t have that luxury: Every conversation runs through an LLM, and the cost depends on length, complexity and the model used. The meter is always running.
“Your cost per customer is no longer fixed,” says Akbasaran. “It fluctuates constantly, meaning your profit margin on any given enterprise client remains genuinely uncertain.”
In one case, Akbasaran saw a single under-monitored product feature generate LLM costs in the six figures annually due to a poor combination of model choice and billing structure. Catching that kind of margin drift requires treating cost monitoring as a core product decision, rather than an infrastructure afterthought.
Actual resolution rates also depend heavily on factors outside the provider’s control, such as the client’s internal data quality and system configuration. Akbasaran notes that this volatility makes value-based pricing difficult to anchor. Companies are forced to price an asset whose ultimate operational value remains unpredictable.
A new framework for corporate experimentation
Addressing these challenges requires a new approach to corporate experimentation, especially in B2B enterprise software, where long sales cycles and large key accounts can easily distort data. To solve this, Akbasaran developed a strategic framework that separates the rigor a test needs from the permanence of the final decision — two questions that product teams often conflate.
“A common mistake is running a light test like pitching a new pricing structure to twenty customers, seeing a positive signal, and rolling it out company-wide,” Akbasaran says. “Twenty customers in B2B tells you almost nothing. But the reverse is also true: teams spend months designing statistically rigorous experiments for low-stakes decisions they could easily reverse in a day.”
The next economic standard
For the broader AI ecosystem, operational discipline is now a baseline requirement. The pricing models dominating the market today were adapted from older software categories and were never designed for the unit economics of LLM.
As the market matures, the industry will have to converge on a new economic standard. “Right now, everyone from early-stage startups to legacy enterprise vendors is working with incomplete information,” says Akbasaran. “The companies that solve this challenge first will gain a significant advantage. The winners will be those who treat this uncertainty as a reason to experiment more rapidly, not those waiting for a playbook to appear.”
The news and editorial staffs of the baiduhai had no role in this post’s preparation.