The brief is the product now.

AI makes it easier to produce a first draft, a working prototype or a research summary. That puts more weight on a different skill: specifying what the work needs to achieve.

The cost of producing knowledge work is in freefall.

Nate Jones called this a two-class system forming inside every knowledge work function: specifiers who define what should happen, and operators who execute with AI. He's right about the pattern. But I think the implications are bigger than a restructuring of roles inside companies. What's forming is an entirely new economy. I'm calling it the Specification Economy.

And the core thesis is simple: the most valuable skill in every knowledge work function just shifted from producing the work to specifying it with precision. The transition is happening now, the barriers to making it are deeply personal, and the cost of waiting is compounding every month.

For decades, the person who could produce the artifact was the scarce resource. The engineer who could write clean code. The copywriter who could craft a campaign. The analyst who could build a model. Production ability was the job.

AI just broke that equation.

When the cost of production approaches zero, production stops being the value. The value migrates upstream to the person who can define, with precision, what should be produced, why it matters, how to validate it, and where the boundaries are.

That's specification. And it's the new bottleneck in every function.

The Evidence Is Already Here

DORA’s discussion of its 2025 findings describes higher AI adoption as associated with both greater software-delivery throughput and greater instability. Faster production is useful, but it still needs review, testing and a clear definition of quality.

A vague brief still needs interpretation. When a person fills gaps with experience, that judgment can be easy to miss. When AI produces the first version, those gaps need to be made explicit and checked.

That's not a software engineering story. It's an everything story.

Every Function. Same Arc.

Software offers a useful early example. Other functions should test where the same tension appears in their own workflows.

In marketing, AI can produce campaigns, write copy, generate creative, and build media plans in minutes. But the CMO who can precisely define the audience, the success criteria, the brand constraints, and the test framework so that AI-generated content actually converts? That person is running circles around the one who's just prompting for "something more engaging."

In finance, AI can build models, run scenarios, and generate reports at a speed that would have been unthinkable two years ago. But the CFO who can define what decision the model needs to inform, what assumptions need stress-testing, and what constitutes signal versus noise? That's where the actual leverage is.

In sales, AI can generate outreach, score leads, and personalize follow-ups at scale. But the CRO who can define the ICP precisely enough and the value narrative clearly enough that AI-assisted outreach actually lands? That's the difference between a pipeline that looks full and one that actually converts.

The question in each case is how well the team defines the work and evaluates the result. More output alone does not answer it.

Why Almost Nobody Is Ready

The barriers aren't technical. They're structural, cultural, and deeply personal.

Thirty years of training in the wrong direction. The professional development system has optimized for specialist production. Go deep. Get credentialed. Master your craft within your lane. Specification requires the opposite: understanding the full workflow, seeing across functions, synthesizing patterns from different industries and business models. The specification economy doesn't just reward depth. It rewards the speed at which you can develop breadth and turn what you learn into something new.

Organizations designed to prevent it. In most enterprises, work is sliced so thin that no single person holds the full context of any meaningful workflow. Strategy defines goals. Operations translates. Functional teams execute fragments. Each handoff loses context. That fragmentation was manageable when humans filled gaps at every step. Now you need someone who can specify an entire workflow with enough precision for AI to execute across those handoffs. That person barely exists in most organizations, because the org was designed to make sure nobody needed to hold that much context.

The leadership identity crisis. This is the hardest one, and the one nobody talks about. Most senior leaders built their careers on pattern recognition and intuition. Being the smartest person in the room. Specification requires the opposite. It requires you to externalize everything you know. Make your intuition explicit. Define "good" in terms a system can evaluate, not something you know when you see it. For a lot of leaders, that feels like a demotion. They've always had the luxury of being vague. Of letting other humans interpret and clean up after them. That luxury is over.

If your people can’t specify the work clearly, AI can help them produce more of the wrong thing. Measure the quality and usefulness of the result alongside the volume.

That's not an AI problem. That's a leadership problem. And the window to fix it is compressing along with everything else.

The Clock Is Compressing

The specification economy doesn't just change what's valuable. It changes the clock.

When production costs collapse, the entire cycle compresses. Idea to working prototype to market test to business result. What used to take quarters now takes weeks. What took weeks takes days. Faster decisions with less room for error. More bets running simultaneously. Shorter feedback loops that require real-time judgment. The ability to kill something that isn't working on Tuesday and redirect by Wednesday.

Most organizations aren't structured for it. Most leadership teams still operate on quarterly planning cycles and monthly review cadences that were built for a world where production took months, not minutes.

Build this skill in the work your team is already doing. Start with a brief, compare the result with your criteria and improve the instructions where the two diverge.

I wrote a full piece on this. It covers the shift in every function, the barriers that make it hard, the jobs question, and what it demands of leaders. If you read one thing this month, make it this one.

Read the full article: The Specification Economy →