The Race for Value Visibility in AI ▍

The Value Visibility Problem

Two developers each spend $25 in tokens with your AI tool. One builds a throwaway survey. The other ships a feature that saves 150 engineering hours. You charge them both $25 because you can't tell the difference.

This is the API boundary blindness problem. And it's breaking pricing for every AI product.

What We Can't See

Your telemetry stops at the API response. The moment your AI generates code or writes an email, you go blind. Whether that output drives $500 or $50,000 of value—you never know.

You built AI systems like infrastructure: count the compute, bill for usage. But AI isn't infrastructure. It's capability. And capability without context is just commodity compute.

You can see token consumption. Can't see if those tokens wrote throwaway scripts or production code. Can't track which AI-generated emails converted to pipeline. Can't tell which analyses changed real decisions.

You measure activity. Your customers measure outcomes. That information gap is where all the pricing leverage lives, and right now it's entirely on their side.

This Is What's Working

Two paths are working:

Path 1: Instrument existing business, then change pricing model

Intercom already had the data—they just didn't structure their business around it. They knew when tickets were successfully resolved. But they still charged per seat.

They extended telemetry past the API boundary into actual workflows. Now they can see the outcome—whether the customer's problem actually got solved. That visibility enabled them to restructure pricing: 99 cents per resolved ticket, not per seat or per customer chat.

That distinction between activity and outcomes is what makes outcome-based pricing possible.

Path 2: Build outcome-focused from day one

Blitzy, started by a former NVIDIA engineer, built tech that understands full codebases and performs major platform rewrites—legacy frameworks to modern equivalents. Their process saves 5x on major platform refactors. (As of writing, they are #1 on SWE-Bench Verified.)

They never charged for tokens. Their go-to-market was structured from the start around the value of an enterprise migration. Their value metric was lines of code. While not completely outcome-focused, they avoided the commodity pricing problem entirely by charging a premium price for a more valuable outcome: a finished platform upgrade. And now, they can't hire fast enough. (There's an unlimited number of enterprises who need to perform major code refactors.)

Either way, start with one workflow

Something with clear outcomes. Code that ships, emails that convert, tickets that resolve. Build telemetry that tracks from AI output to business result.

You don't need perfect attribution. You need directional signal, like Blitzy has. Can we distinguish code that ships from code that gets deleted? Emails that convert from emails that don't? That's enough to start pricing on something closer to value.

Why This Matters Now

Your current pricing is temporary. Every day you operate blind to outcomes is a day competitors might be building visibility.

The first vendor in your category to see value clearly will price accordingly. They'll capture more of the value they create. You'll be stuck defending usage pricing against customers who know exactly what they're worth.

The vendor who can distinguish a $500 survey from a $50,000 feature wins the market. Everyone else competes on token prices.

In an upcoming post I'll write about how verticalized SaaS vendors are uniquely positioned to win in outcome based pricing.