The brief is the product now.

Cursor hit $500 million in annual recurring revenue with 150 people. Midjourney hit $500 million with fewer than 170. Lovable crossed $100 million in eight months with 45 employees. A quarter of Y Combinator's latest batch shipped products with codebases that were 95% AI-generated.

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

Software engineering is 18 months into this transition. Individual developers report feeling more productive than ever. But Google's 2025 DORA report found that organizational delivery, the thing that actually matters, stayed flat or got worse. More output. Same or lower quality. More rework.

The gap between individual productivity and organizational results isn't a mystery. It's a specification gap. When humans handled production, they were quietly interpreting vague requirements and filling gaps with domain knowledge. The engineer didn't just write code. She read a loose spec and thought, "they probably mean this, and they haven't considered this edge case, so I'll handle it." That invisible judgment layer disappeared the moment AI took over production. Now every gap in the specification flows straight to the output. And the output looks confident and polished, which makes the problem harder to catch, not easier.

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

Every Function. Same Arc.

Engineering is the canary. Every other commercial function is on the same curve, just earlier on the timeline.

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 pattern is identical in every case. The ability to produce becomes abundant. The ability to specify becomes scarce. And the gap between the two is where organizations are bleeding value right now without realizing 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 with precision, AI makes them faster at producing the wrong thing. More output. Same quality thinking. No improvement in outcomes. The specification gap is the most expensive problem in your organization right now, and most leaders can't see it because the output looks so good.

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.

And here's the part that should create real urgency: the specification economy isn't arriving in five years. Engineering is already there. The other functions are months behind, not years. The people building specification capability now will be positioned for the roles that emerge on the other side of this transition. The people waiting won't.

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 →