In their book Prediction Machines, economists Ajay Agrawal, Joshua Gans, and Avi Goldfarb offer a clarifying framework: AI is prediction technology. Machine learning takes inputs and maps them to outputs: predicting what comes next, what will resonate, what patterns the data suggests. As this capability improves and scales, prediction becomes cheap.
The Changing Economics of Prediction
The insight that follows is economic: when an input gets cheap, value shifts to its complements.
When prediction was expensive, organizations paid handsomely for it. Entire industries existed to forecast demand, model scenarios, anticipate trends. Now a foundation model can do this in seconds, essentially for free to the user.
So where does value migrate? According to Prediction Machines, value moves to things prediction cannot replace, and to the infrastructure required to use prediction safely at scale:
Good Data: Not raw data, but structured, contextualized, well-organized data. The machine is only as good as what you feed it.
Judgment: What do we want? What's acceptable? What trade-offs will we make? Prediction tells you what's probable. Judgment tells you what's worthwhile.
Compliance: When prediction becomes cheap, decisions scale. When decisions scale, mistakes scale too. What governance protocols dictate how all of this is handled? Compliance is judgment again, really.
Action: Execution. And yet, with Gen AI models, actions appear to become increasingly commoditized. AI can draft, design, and deploy in an instant. But execution without strategy is noise.
The Rare Skills
Prospero sees the importance of judgment, again and again. We also see the need for strategy.
Judgment is the irreducibly human input. Judgment is the human choice itself.
Strategy is the layer built on judgment, informed by prediction. The objective, the direction, the decision. Strategy is synthesis: here's what we know, here's what we predict, here's what we've decided, here's what we'll do.
Agency Corollaries
The old model, where agencies sold execution hours and called it strategy, is dying. You don't need a hundred people to produce content.
You need judgment and strategy. This work demands a partner you can trust.
Reputation work becomes more strategic, not less. When anyone can publish anything, the signal-to-noise problem explodes. Credibility becomes the scarce resource. Earning it through judgment, consistency, and trust is the game.
Prospero exists for this moment. We are a judgment and strategy provider. This is where we see the real value to add in an AI and hyper-digital world.