Are You Buying AI Per Seat Or Per Token? That's the Wrong Question. Ask This Instead.

Executive summary

The video argues that enterprise AI pricing debates—per seat, per token, or per job—often focus on the wrong issue. AI is a new economic input: intelligence. The central business challenge is not simply buying it cheaply, but organizing people, workflows, and partnerships to turn it into measurable customer and company value.

Pricing should reflect how value is created. Token pricing can help control and forecast usage, while outcome-based pricing may suit clearly defined, closed-loop tasks. More complex work may require consulting, embedded engineering, or deeper partnerships. No single commercial model fits every company; the right approach depends partly on how capable the organization is at applying AI.

Business strategies and operating playbooks

• Start with the business outcome, not the pricing unit. Ask what customer or company result AI enables and whether that result can be measured. Tokens measure model usage, not the value of the work produced.
• Build value through a frontline discovery loop:
  1. Identify a recurring customer or operational problem.
  2. Use AI to gather and compare relevant information.
  3. Apply domain expertise to validate patterns and determine root causes.
  4. Decide what AI can handle, where human judgment is needed, and when to involve a specialist or customer.
  5. Repeat the work and look for a service or product that can be packaged and sold.
• Treat deployment as an end-to-end process. Identify an opportunity, scope it, design and build a system, test it, integrate it with company data and processes, and support its use in daily operations.
• Balance experimentation with execution. A demo-ready model does not mean the organization is ready to use it. At the same time, companies should not rush to turn every early use case into a finished product. Frontline learning and disciplined implementation are both necessary.
• Test growth opportunities, not only cost savings. One suggested experiment is to fund a project the company previously lacked the capacity to pursue, then measure whether AI-enabled productivity helps generate new revenue.
• Choose pricing to match the work:
  • Per token or usage: Useful for budgeting and experimentation, but does not guarantee useful output.
  • Per seat or license: Familiar procurement models, but may not reflect business outcomes.
  • Per job or outcome: Potentially suitable for well-defined tasks with clear completion criteria.
  • Consulting, embedded teams, or joint ventures: May be more appropriate when success depends on process changes, specialist knowledge, and ongoing collaboration.
• Make employees “intelligence managers.” People should decide where to apply AI, how to evaluate its output, and when to shift to human work. This role can exist before someone receives a formal management title.
• Support and incentivize frontline ownership. Effective AI adoption depends on employees having tools, motivation, relevant expertise, and a sense that the company supports their initiative.

Examples cited

• Cedar and its AI agent, Cora: A patient can call about a medical bill, receive an explanation, confirm insurance, and pay during the same interaction—even at 2 a.m. The example illustrates a closed-loop service in which the patient and provider may care more about resolving and paying the bill than about token consumption. The speaker does not know Cedar’s contract terms.
• Equipment-maintenance scenario: An engineer uses an agent to review maintenance records and past incidents, then applies domain knowledge to distinguish similar-looking problems and identify root causes. Repeated work could develop into ongoing customer consultations or a new product.
• OpenAI Deployment: Described as working with business and industry teams to identify opportunities and design, build, test, and implement systems tied to client data and processes.
• Anthropic and a venture with Blackstone, Hellman & Friedman, and Goldman Sachs: Presented as examples of applying AI with businesses such as mid-sized manufacturers and regional health systems.
• OpenAI and Thrive Holdings: The video says OpenAI acquired a stake in Thrive Holdings and planned to embed research, product, and engineering teams in its companies, initially focusing on accounting and IT services.
• Sierra: Cited as exploring outcome-based deals in which the provider and buyer agree on a price for a defined result. The speaker suggests this is easier for closed customer-service interactions than for complex transformations.
• Shopify: COO Jess Hertz is cited as describing AI returns while engineers use different models for different circumstances and costs are managed. The speaker says Shopify kept staffing roughly flat, postponed hiring for about two years, and grew revenue by 34% during that period.
• Block: The video says Block reported 25% gross-profit growth while significantly reducing its team and putting AI at the center of its business. No specific period or headcount figure is provided.

Metrics and caveats

• Shopify: Revenue growth of 34%; hiring postponed for approximately two years; staff size described as remaining roughly the same.
• Block: Gross profit increased 25%; the team was reduced significantly, with no quantified staffing change given.
• No AI pricing rates, margin figures, ROI targets, or implementation timelines are specified. The examples and figures are presented as claims in the video, not independently verified here.

Management takeaways

• CFOs and CEOs should evaluate AI investments based on the company’s ability to create value with them—not only on license, token, or usage costs.
• Budget controls can enable experimentation, but excessive focus on limiting spend may prevent companies from developing valuable use cases.
• Before purchasing more AI, assess whether the organization has the expertise, processes, and employee support to apply it effectively.
• For individuals, a practical career opportunity is to learn how to identify valuable uses of AI, manage its application, and show how the work improves customer outcomes or business performance.

Presenter: Nate B. Jones

Organizations and people mentioned: Cedar, Cora, Sierra, OpenAI Deployment, Anthropic, Thrive Holdings, Blackstone, Hellman & Friedman, Goldman Sachs, Shopify, COO Jess Hertz, and Block.