Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Christina Cacioppo says Vanta avoids marking up model tokens because users quickly compare it to Claude and ask why the markup is better. Vanta is absorbing token costs, accepting lower margin, and focusing on getting more customers into agent workflows before sorting out pricing later.

Related Insights

Budgeting for AI is difficult because the utility-based, per-token pricing model is not viable or scalable for business departments like marketing and sales. This system is a temporary phase; expect AI providers to shift toward more predictable, outcome-based pricing models as the technology matures.

Pure value-based pricing (e.g., per seat) fails for AI products due to unpredictable token costs from power users. Vercel's SVP of Product advises a hybrid model: one metric aligned with value (like seats) and another aligned with cost (like token usage) to ensure profitability.

Many AI coding agents are unprofitable because their business model is broken. They charge a fixed subscription fee but pay variable, per-token costs for model inference. This means their most engaged power users, who should be their best customers, are actually their biggest cost centers, leading to negative gross margins.

Moving from simple chatbots to autonomous agents creates a massive cost increase. Agents consume 5 to 30 times more tokens because they operate in loops, with each task involving 10-20 separate model calls that carry extensive history, instructions, and tool definitions, rapidly compounding costs.

Unlike traditional SaaS where unit costs stay relatively flat, AI features incur per-interaction token expenses. Product and engineering teams must track costs down to workflows and users in the codebase to prevent uncontrolled margin erosion.

In a crowded market where startups offer free or heavily subsidized AI tokens to gain users, Vercel intentionally prices its tokens at cost. They reject undercutting the market, betting instead that a superior, higher-quality product will win customers willing to pay for value.

A model with a low per-token price can be more expensive if it's inefficient, verbose, or requires multiple attempts ('overthinking'). The actual invoice depends on the total tokens needed to complete a task, making token efficiency a hidden multiplier that savvy enterprises are now tracking to determine the true cost.

Customers are intimidated by token-based pricing. Offering a flat-fee "unlimited agents and usage" package removes this friction. In reality, clients rarely need more than a few well-configured agents, making the model profitable and simple to sell by focusing on value instead of usage.

A few months ago, the fear was AI replacing SaaS businesses. Now, the pressing issue is managing massive AI bills. This has elevated 'token economics'—optimizing costs by using different models for different tasks (model routing)—from an advanced technique to a non-negotiable, table-stakes practice for any serious AI implementation.

Sticker price per token is a misleading metric for AI models. A cheaper model may require more retries or reasoning, making it more expensive overall. The true metric is 'cost per accepted task,' which accounts for total resources needed to get a reliable, usable result, providing a true apples-to-apples comparison.