Summaries > AI > Ai > Are You Buying AI Per Seat Or Per Token? That's the Wrong Question. Ask This Ins...

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

https://www.youtube.com/watch?v=RvnQZQAHUDc

TLDR AI’s value depends less on token or seat pricing than on how well people direct it toward useful business outcomes. Companies should empower employees to identify the right tasks, verify AI’s work, and involve humans when needed, while experimenting with partnerships, embedded engineers, and outcome-based pricing to turn AI capability into measurable customer and business value.

Key Insights

Start by Defining the Business Outcome

Before choosing an AI tool or pricing model, identify the customer or business result you want to improve. Token usage measures computing activity, not whether the work created value or delivered a return. For example, a healthcare agent may be valuable because it helps patients understand bills and complete payments—not because of how many tokens it uses. Set a clear outcome and decide how you will measure it.

Match the Pricing Model to the Work

Per-token pricing can help manage usage, and per-seat pricing may fit familiar software arrangements, but neither automatically reflects business impact. Outcome-based pricing can work well for discrete, measurable services, such as resolving customer-support requests or collecting a payment. It is harder to apply to complex work, such as redesigning a manufacturing service program, where results have many causes. Choose a model that fits how clearly you can define and measure the outcome, and keep experimenting as you learn.

Put People in Charge of Applying AI

AI creates value when people decide what to investigate, what to delegate, and when human expertise is needed. Employees should check AI-generated findings against their domain knowledge, investigate important patterns, and intervene when a task requires judgment or direct customer engagement. In equipment maintenance, for instance, an AI agent can flag recurring failures, while an engineer verifies the pattern and determines its cause. Treat employees as managers of AI, not merely users of a tool.

Build Capability Around Real Workflows

Organizations that are less fluent in AI may benefit from embedded specialists or forward-deployed engineers. These teams work with business leaders and frontline employees to identify valuable opportunities, then build and test solutions using the company’s actual data and processes. This practical support helps close the gap between purchasing AI and putting it to effective use. As teams gain experience, they can take greater ownership of AI adoption.

Learn Through Services Before Scaling Successful Solutions

Companies can deliver AI-enabled work as a service while learning which problems recur across customers. Moving too quickly to a standard product may limit insight into different needs, while staying in custom service work indefinitely can make successful solutions harder to scale. Use early projects to discover repeatable tasks, validate results, and understand what customers value. Then turn proven approaches into products or offerings when the evidence supports it.

Consider Deeper Partnerships When Value Is Shared

Some AI opportunities may require a closer commercial relationship than a conventional software purchase. Embedded teams, outcome-based agreements, or joint ventures can combine an AI supplier’s expertise with a company’s operational knowledge. These arrangements may help share in the value created, but they also involve greater commitments and can be difficult to measure fairly. Define responsibilities, success measures, and risks before entering a deeper partnership.

Track Productivity Gains as Business Results

Monitor whether AI is improving meaningful outcomes, not just reducing costs or increasing usage. Look for evidence such as faster service, better customer experiences, increased revenue, or capacity freed for work the company previously could not resource. Give frontline teams the support and authority to identify useful applications and share what works. When leaders connect those improvements to business results, they can make better decisions about investment and pricing.

Questions & Answers

Why don’t per-token and per-seat pricing fully capture the value of enterprise AI?

Per-token pricing measures compute, not useful thinking or business impact; per-seat pricing mirrors SaaS but may not reflect the outcomes AI creates. In both models, companies still need to turn AI capability into valuable work.

What does it mean for employees to become “managers of intelligence” or “managers of AI”?

Employees must decide what work to assign to AI, evaluate its results, adapt processes, and recognize when human expertise is needed. Their judgment helps direct AI toward meaningful outcomes.

How can companies turn purchased AI into working systems?

Forward-deployed engineers can work with business leaders and frontline teams to identify opportunities, then build, test, and deploy solutions within the company’s actual data and processes.

When should AI companies provide services, build products, or form deeper partnerships?

Companies should balance customer-facing AI services with turning repeated successes into products, since moving too quickly to either model can limit learning. For some opportunities, deeper partnerships—such as embedded teams, outcome-based arrangements, or joint ventures—may be appropriate, though they require significant commitment and measurable outcomes can be difficult to define.

When does outcome-based AI pricing work best, and when is it difficult?

It can work well for discrete, measurable services, such as customer support or an agent that explains a medical bill and collects payment. It is harder to apply to complex work, such as redesigning a manufacturing service program, where results are less straightforward to measure.

What do Shopify and Block illustrate about AI-driven productivity?

Shopify COO Jess Hertz says engineers should use the best models for each situation while spending is monitored. Shopify kept head count flat as revenue grew 34% over roughly two years. Block reported 25% gross-profit growth after reducing its team and placing AI at the center of its business, though those results will need to be assessed over time.

What should company leaders measure when assessing AI’s value?

Leaders should look beyond AI’s price and cost reduction. They can test whether AI frees capacity for revenue-generating work that was previously impossible to resource, then track the resulting customer and business value.

What helps companies that are not yet fluent in AI capture a return on investment?

They may need partnerships, embedded engineers, or consulting support to build the capability to apply AI to important problems. They also need motivated, supported employees who can connect productivity gains to customer value and business results.

Summary of Timestamps

Enterprise AI pricing often relies on familiar models: charging per token to meter compute or per seat, as with SaaS. But neither directly measures the economic value of AI. Tokens can help control spending, yet they do not guarantee useful results or return on investment.

AI agents can deliver intelligence directly to customers, making outcomes more important than the underlying token count. Cedar’s healthcare payments agent, Cora, for example, explains medical bills, confirms insurance, and helps patients pay—services hospitals may value more than the compute required to provide them.

As AI becomes more capable, employees at every level become managers of intelligence. They must decide what work to assign to AI, evaluate its output, adapt processes, and recognize when human expertise is needed. AI creates value through these human decisions, not through automation alone.

Many companies struggle to earn returns because they lack the know-how, motivation, alignment, or confidence to use AI effectively. The central challenge is not simply choosing a pricing model or asking AI to solve a problem; it is setting an ambitious strategy that helps employees direct AI toward meaningful business outcomes.

An equipment company illustrates the role of human judgment: an engineer might use an agent to compare maintenance records and find recurring failures, then apply domain knowledge to verify the patterns, investigate causes with customers, and decide when to bring in a person.

Forward-deployed engineers help organizations turn AI purchases into practical systems. Working with business leaders and frontline teams, they identify opportunities and build, test, and deploy solutions within real company data and processes. OpenAI and Anthropic are pursuing this approach, including Anthropic’s partnership with Blackstone, Hellman & Friedman, and Goldman Sachs.

Companies need to balance delivering AI as a service with turning repeated successes into products. Moving too quickly toward either model can limit learning. For instance, repeated success in preventing equipment failures might lead to a new offering grounded in customer relationships, operational knowledge, and credible recommendations.

Some opportunities may call for deeper commercial partnerships rather than standard software purchases. OpenAI’s investment in Thrive Holdings, whose embedded teams initially focus on accounting and IT services, illustrates how AI suppliers might contribute expertise and share value through outcome-based pricing or joint ventures. These arrangements can offer gains but require substantial commitment, and outcomes may be difficult to measure.

Successful adoption depends on motivated, supported frontline teams who can recognize when AI is useful and apply it to customer needs. Outcome-based pricing may fit discrete services, such as customer support or Cedar’s bill-explanation and payment collection process, but it is harder to measure in complex work like redesigning a manufacturing service program. Per-token pricing will remain, but companies can also test models tied to broader value.

There is no single commercial model that suits every company; the right arrangement depends partly on its AI fluency. Shopify COO Jess Hertz says engineers should use the best models for each situation while she monitors spending. Shopify has kept head count flat as revenue grew 34% over roughly two years, which she views as evidence of measurable AI-driven productivity.

Block has also reported 25% gross-profit growth after reducing its team and placing AI at the center of its business, although the longer-term results remain to be seen. These examples show why companies should track whether AI productivity translates into durable business performance.

To capture clear ROI, organizations need people who can apply AI to important problems and connect productivity gains to customer value and business results. Less AI-fluent firms may benefit from partnerships, embedded engineers, or consultants. Employees can also lead by becoming managers of AI, demonstrating its value and helping leaders apply it strategically.

Leaders such as CFOs and CEOs should assess their organization’s ability to create value—not just AI’s price. Rather than focusing only on cost reduction, they can test whether AI frees capacity for revenue-generating work that previously lacked resources, then measure the results. The speaker closes by inviting people to share their AI experiences and views on per-token versus outcome-based pricing.

Related Summaries

The Newsletter Is Dead. Hereâ€™s How To Survive AI ...

Vercel CEO: How 1 Person + AI Agents Could Build a ...

Why Cem Karsan Is Saying “Let’s Go” as Everyone Els...

What Claude Code Can Do That You Haven't Tried - Gu...

How to generate outlier investment ideas – intervie...

I Ran 1,000 Email Deliverability Tests. Here's What...

Menu