AI Pricing: The Shift Towards Outcome-Based Models in Enterprises

AI pricing is undergoing significant change as enterprises transition from paying for effort to paying for outcomes. Startups that develop an outcome-pricing layer, especially in India’s inference stack, have the potential to lead the next phase of AI adoption.

The Current State of AI Pricing

Presently, AI technology is usually sold based on the amount of effort the model expends (measured in tokens), rather than the results it produces for users. This mismatch is leading to increasing discrepancies in expectations among enterprises globally.

As some of the largest AI firms move toward initial public offerings (IPOs), their claims suggest that AI solutions will replace parts of the human workforce, necessitating a resolution to this paradox sooner rather than later.

Concerns in the Enterprise Sector

During an event in late May, OpenAI’s CEO, Sam Altman, acknowledged that the costs associated with tokens have become a significant concern, stating that customers express issues related to budget management. Customers have conveyed that their organisations exhausted their entire budget early in the year, prompting a call for more cost-efficient solutions. Altman mentioned that cost is now the second-most frequent complaint he receives from enterprise clients, following only the desire for simpler workflows.

Goldman Sachs predicts that by 2030, the use of agentic AI could increase token consumption by 24 times, reaching 120 quadrillion tokens per month. The trend is evidently escalating.

Budgeting for AI: Insights from Practice

On a smaller scale, operators managing multiple AI automations noted that budget planning often focuses on tools and integration without factoring in token usage. After three weeks of use, the focus shifted from assessing time savings to evaluating whether using AI is indeed less expensive than hiring staff.

Payment Models: Humans vs. AI

In contrast to AI, where costs are calculated based on effort, human professionals are compensated based on outcomes. When a founder hires a contractor to draft legal documents, they pay a flat rate for the final product, not for the attorney’s time spent reviewing legislation. This reflects a mature professional services market, where prices are set by project completion, procedure, or transaction.

However, the current token billing approach reverses this system. Clients end up paying for the deliberations, retries, and inefficiencies related to AI processes, regardless of whether these lead to valuable results.

Without incentives to operate efficiently, AI providers may even have reasons to encourage excessive usage. While checking for code errors, AI could unnecessarily analyse the entire codebase rather than focussing just on current issues.

The prevalent effort-based pricing is typical of early-stage service markets, where measuring output is simpler. True outcome pricing emerges when providers have confidence in their ability to deliver, a state that AI has yet to achieve. However, the market demands this evolution.

Distinct Features of Token Billing

Token pricing displays three structural distinctions absent in previous SaaS and cloud models:

• Firstly, there is no consistent unit of measure. SaaS models relied on user seats; cloud services were based on compute hours. Tokens, however, lack a stable cost curve or a predictable consumption pattern; a minor change in prompts can result in markedly higher costs.
• Secondly, the absence of departmental ownership complicates budgeting. AI usage spans various business functions, making it impossible for finance departments to allocate expenses to specific profit and loss statements. A financial analyst familiar with every line item struggles to decipher AI-related invoices.
• Lastly, there is no common metric of value. Sales teams track qualified leads while support teams monitor resolved tickets; in contrast, all a token signifies is that usage occurred, without providing insight into whether the task contributed to achieving a goal or offered value.

Real-world Impacts on Corporations

Uber exemplifies the challenges inherent in this pricing model; 95% of its engineering team utilises AI tools monthly, yet the Chief Operating Officer, Andrew Macdonald, indicated difficulty in linking AI spending to tangible improvements in their products. High internal usage has not translated to identifiable outcomes.

In contrast, companies demonstrating high AI engagement have seen their revenues double since 2023, whereas those with low engagement have seen stagnant growth. The solution lies not in reducing spending, but rather in applying the same level of scrutiny to AI expenditures as firms do for hiring and vendor management, necessitating precise outcome metrics, which token billing does not supply.

Companies Opting for Control Over Flexibility

The trend of rationing AI resources has emerged, affecting major organisations. In April, Uber’s CTO disclosed that the company had depleted its entire AI coding budget for 2026 in just four months. Adoption rates for Claude Code surged from 32% to 84% among its engineering staff within a mere three months.

Individual billing for engineers varied significantly, with expenses reaching between £500 to £2,000 monthly. The CTO personally accrued £1,200 in token costs during a two-hour internal demonstration. By May, Uber implemented a cap of £1,500 per month for each engineer on AI coding tools.

Similarly, Microsoft discontinued the majority of Claude Code licenses for its experiences and devices division, which includes Windows, Microsoft 365, Outlook, Teams, and Surface. This decision redirected thousands of engineers towards GitHub Copilot CLI by the end of June. While the official reason cited was for tool integration, the timing suggests cost management was equally crucial. Around this same period, GitHub Copilot transitioned from a flat monthly fee structure to token-based billing, causing users to run out of credits sooner than expected.

This transition reflects the pattern found in every IT sector; informal shadow IT becomes regulated governing IT, with SaaS oversaturation leading to vendor consolidation. AI is currently entering a phase of consolidation, progressing more quickly than past IT cycles.

Founders should prepare for enterprise customers to shift their inquiries from “what can your AI do” to “what is the cost per outcome delivered” within the coming year. Companies providing clear answers to this question will secure deals, while those simply quoting token rates may struggle.

Developing the Outcome-Pricing Model

The market is in the midst of a transformation towards outcome pricing. Leading enterprise vendors are adapting their strategies.

Salesforce has introduced Agentforce, now pricing AI-initiated actions based on agentic work units linked to workflow completion and case resolutions. Zendesk has committed to an outcome-based pricing structure for its AI agents, with CEO Tom Eggemeier asserting that as AI agents take on more workload, customers should anticipate prices that reflect outcomes achieved.

Other companies, such as Sierra, charge for each customer support conversation that gets resolved, providing no fees for unresolved cases. Fini has set a fee of £0.69 for every AI-resolved ticket. ServiceNow, UiPath, and Adobe are also transitioning towards hybrid pricing models that combine subscriptions, consumption, and outcome elements.

This transition is not a simple task. It necessitates the development of three essential infrastructures, which are largely lacking at present:

• First, a clear definition of outcomes, aligning with the standard measures for tasks performed by sales agents, support agents, or code agents.
• Second, a method for verifying outcomes, confirming whether the agent actually resolved a task or merely marked it as complete.
• Third, acceptance of risk by the seller, who must manage token variance, necessitating smart model routing or healthy margin buffers to cover worst-case scenarios.

Inference Stack: The Key to Outcome Economics

To effectively implement outcome pricing, control over the cost-per-outcome calculations is essential. This control resides in the inference layer rather than the application layer. Indian infrastructure companies are making significant strides in this area.

One Indian AI cloud provider reports processing 40 billion tokens daily through its inference platform. Indian businesses are experiencing inflated token costs due to running production workloads on shared infrastructure designed primarily for exploratory testing.

By migrating their production workloads from experimental APIs to a dedicated inference stack utilising open-weight models, these enterprises can significantly reduce their monthly AI costs. They have the potential to decrease expenses from £75,000 to much more manageable figures.

Paying for Outcomes: AI’s Biggest Opportunity

Paying for outcomes is a significant development in the AI sector. The cost associated with specific tasks can be significantly reduced; for instance, expenses can drop from $81,000 to $9,000, marking an 88% reduction without compromising the task’s capability.

Understanding the Directional Reading

The directional reading remains clear. Not every task necessitates the most advanced models. For example, summarising a meeting doesn’t require GPT-5.5 Pro, and not every code review needs the Claude Opus 4.8. The intelligent routing of models alone can drastically lower AI costs, enabling outcome-based pricing to function effectively within the seller’s margin.

India’s Growing Innovation Landscape

India stands poised as the second-largest inference market globally. The global inference market is anticipated to surpass training expenditures by 2027, with expanding opportunities as this trend becomes more mainstream.

The Indian SaaS Playbook

The framework for Indian SaaS over the past decade has been based on improved unit economics paired with competitive pricing. Factors contributing to this success include lower engineering costs, stricter capital management, and a culture focused on profitability from early stages. This same advantage can drive the shift to outcome-based pricing.

Competitive Advantage for Indian Companies

A vertical AI firm in India that manages to price by outcome, route intelligently between models, and account for token variance will surpass a global competitor that still relies on traditional per-token pricing. Indian founders are accustomed to this mindset, and their customers expect a similar approach.

Emerging Trends in AI Startups

Reports from May 2026 indicate that Indian AI startups are transitioning from seat-based models to flexible pay-as-you-go and outcome-linked structures, responding to customer demands for adaptation. The presence of 2,117 Global Capability Centres (GCCs) in India employing 2.36 million people and generating $98.4 billion in revenue underlines this shift. Over 1,200 of these GCCs have integrated AI and machine learning capabilities, with a skilled AI workforce of approximately 250,000 professionals.

The Challenges Faced by Buyers

These companies are grappling with similar cost issues as Uber and Microsoft concerning token billing, while also managing the added complexities of data privacy and compliance that turn every cross-border prompt into a risk.

Exploring Category Evolutions

Diving deeper, several specific categories are evolving, such as vertical agents in fields like legal, accounting, customer support, performance marketing automation, asset generation, system design, design verification, and code review. Each domain has a distinctive outcome unit, with buyers already familiar with outcome-based vendor pricing.

The Future of Business Costs

In the new landscape, tokens are becoming a crucial component of business expenses, alongside personnel and suppliers. They are unlikely to disappear. However, the current billing approach is derived from the cloud era, which does not align well with the economics of AI.

The upcoming leaders in the industry will not merely be those with the cheapest tokens or the most extensive models. Instead, success will come to those who can effectively price outcomes on both the buying and selling sides.

The post AI’s Biggest Opportunity Lies In Paying For Outcomes, Not Tokens appeared first on StartupSuperb Media.