# Thoughts on Outcome-Based Pricing In September 2024, Zendesk made an announcement that signaled a fundamental shift in how AI services are monetized: they would only charge customers when their AI agents successfully resolve customer issues without human intervention. This move represents the emergence of outcome-based pricing in the AI industry, where vendors align their revenue directly with the value delivered to customers. In this essay, I’d like to discuss some of the possible implications of this shift. Software pricing has evolved significantly over the years, reflecting broader shifts in how software is developed, delivered, and consumed. In the early days, perpetual licenses dominated. Companies would pay an upfront fee to own a software product outright, often alongside annual maintenance fees for updates and support. This model aligned with the era of physical software distribution, where products were shipped on disks and installed on local machines. The rise of the internet and cloud computing ushered in subscription-based pricing. Software-as-a-Service (SaaS) allowed customers to access software via the web for a recurring monthly or annual fee. This shift lowered barriers to entry, enabling companies to attract a broader customer base. In the past decade, usage-based pricing models have gained traction. This approach ties cost more closely to value by charging customers based on metrics like data usage, API calls, or active seats. Today, an estimated 70-80% of software companies rely on usage-based pricing. Over the next few years,  however, it is possible that the advent of AI will usher in a new pricing model: outcome-based pricing. Under an outcome-based model, customers only pay when specific, measurable results are achieved. For AI-powered customer service platforms, for example, this means charging per successful resolution rather than per user or consumption. This approach aligns software pricing even more closely to the value it provides. It’s worth noting that software itself has become more powerful over time. The old perpetual-license products of the 90s couldn’t match the sophistication and features of modern SaaS tools. With usage-based and outcome-based billing, the connection between price and value has tightened significantly. In theory, that should boost both the value customers receive and the revenue software companies capture. ![[CleanShot 2025-01-12 at [email protected]]] _Newer business models are better at capturing value created. Outcome-based pricing is linearly correlated with value created_    Yet competition can undermine margins, even with outcome-based pricing. If Zendesk decides to charge per resolution, a competitor could simply offer a lower rate. The most successful companies would be the ones able to provide resolutions just slightly more effectively, from a cost perspective, than the competition. Meanwhile, rival platforms might go the opposite route, reverting to flat-rate subscriptions in an attempt to win customers who prefer predictable costs. There’s another inherent problem in outcome-based pricing: not all outcomes incur the same costs. For instance, a successful customer service resolution may, in some cases, only require emailing the customer a link to reset their password; in other cases, a resolution may require dozens of back-and-forth messages and significant test-time compute. From the client’s perspective, both are “successful resolutions,” but for the AI provider, one scenario may be highly profitable while the other leads to a loss. An interesting side-effect of outcome-based billing is that it attaches specific price tags to company goals. Need a support ticket resolved? That’s $1. Want a feature coded? That’s $50. This could make businesses more capital-efficient, as they’ll see exactly what outcomes cost. For software providers, this can be either a blessing or a curse, depending on the value they’re able to deliver to customers—and how tightly that value is tied to their bottom line. So why switch to outcome-based pricing? The main reason may be that it’s increasingly hard to measure and bill for AI’s unseen processes. If AI services grow more “invisible”—with interactions occurring over email, for example—tracking usage becomes even trickier, which makes billing by results more appealing. Another major driver is the tight alignment with customer success. When vendors profit only from measurable results, they showcase real confidence in their product and lower perceived risk for clients. This outcome-based commitment also spurs deeper insight into what truly matters to customers, often sparking new features or solutions tailored to the most valuable use cases. Still, not every industry can easily adopt this model. Cybersecurity is one example: increased security can dissuade attacks, but it’s possible you wouldn’t be attacked even without the additional security—it can be very difficult to tell. One creative solution may be found in the traditional business model for Chinese acupuncture. In many small villages, acupuncturists were paid a small monthly fee and served as de facto primary care providers. When a villager fell ill, they stopped paying the acupuncturist until they recovered—an arrangement that incentivized the acupuncturist to keep everyone healthy in the first place. Similarly, a “reverse-outcome” model for cybersecurity could involve a monthly usage fee that gets discounted whenever a successful attack occurs. Outcome-based pricing may well be the pinnacle of value alignment. For companies like Zendesk, it offers a competitive differentiator, but it’s also a gamble. Its success depends on how well organizations can define, track, and monetize the outcomes their software delivers—and whether customers embrace this new way of paying for results.