Decisions, Not Documents

AI can now think faster than we can read. The standard unit for AI is a million tokens (roughly a million words), and the normal price for this is a few dollars. But a million words is the equivalent of The Hobbit, The Lord of the Rings, and The Silmarillion—combined. But while thinking has become cheap and abundant, our ability to process and act on it has not.

This changes everything. One hour of a strategist’s time used to buy one hour of human thinking. Now it buys a library. But reading the library is still slow. The bottleneck has shifted. Intelligence is no longer scarce. Strategic discernment is.

We built our current strategic workflows around this older reality. That wasn’t arbitrary—it was optimized for an all-human workflow. Humans read slowly, synthesize contextually, and need to communicate clearly across teams. So we shaped our tools and timelines to fit those limits. Now that constraint is changing, and we need to flip the stack.

How We Think Today (Because We Had To)

Human research is slow—but thoughtful. We often blur the lines between gathering information and shaping understanding. That blend can be powerful: reading, reflecting, and refining as we go. But it also means that every piece of research comes at a cost.

Because it’s expensive, we treat the results as precious. We write them down—documents, memos, notes, decks—so they can be preserved and reused. But even the modest volume of human-generated insight quickly exceeds what others have time to read. So we summarize, distill, and package. We craft whitepapers and reports meant to carry meaning across teams, time zones, and use cases.

And because all that compression takes so much effort, we only do it once. Which means the final product has to serve many audiences, many needs. It becomes generic—not because we lack intent, but because we can’t afford to tailor it.

This is the legacy mode: slow, handcrafted thought designed to travel. It was necessary when the costs of reading and writing were high. But it also locked strategy into a world of static artifacts and one-size-fits-all thinking.

Flip the Stack: Decision-First Strategy

In a world of abundant AI, we no longer need to focus on gathering information - we can have all the research we want, almost for free. Instead, we need to focus on why we're doing this in the first place.

Don’t gather data and then ask what it means - there's an infinite amount of data. Flip it. Start with: What are we trying to decide? What matters for that decision? What are the criteria, constraints, and trade-offs? What kinds of evidence or comparisons would clarify the path?

Once those questions are clear, we can use AI to generate precisely what we need:

Targeted summaries focused on the key variables.
Alternative strategies or paths.
Comparisons of options with pros and cons.
Critiques or implications under different assumptions.

It’s not about using AI to jump to a hasty answer. It’s about shaping a set of high-quality options, structured around a meaningful question. You’re not trying to eliminate ambiguity—you’re trying to make it visible and manageable.

When thought is cheap, the most valuable act is no longer to gather, but to frame. Strategic leverage moves upstream—from the act of reading to the act of asking.

What Better Thinking Looks Like

We all started interacting with AI via chat. It felt revolutionary—ask a question, get a fluent response. But chat moves at human speed. It keeps us in the loop, shaping and steering, one turn at a time. That’s great for exploration—but slow for depth.

The next generation looks different. Tools like Perplexity and emerging Deep Research features now let you launch a single question and return minutes later to a structured, multi-page report. They search dozens of sources, synthesize key findings, and format the result like a briefing document. Why does that work? Because they’ve scaffolded a common task: gather data from across the web and summarize it clearly.

That’s just the beginning. What’s coming next are workflows tailored not to generic research questions, but to the specific decision you need to make—complete with your framing, your constraints, your trade-offs. If you can clarify what you’re deciding, the system can increasingly do the rest.

This shift isn’t about better answers. It’s about building workflows that think.

Start Here: Design for Decision, Not Data

Imagine you had McKinsey building the perfect custom report for your next decision—not to tell you what to do, but to lay out the critical trade-offs, options, unknowns, and decision factors. What would be in that report? What sections would it include? What would the table of contents say? What assumptions would be tested?

Use that mental model as your starting scaffold. You’re not prompting AI from scratch. You’re designing a decision brief. Use it to focus each AI task: explore alternatives here, summarize stakeholder positions there, generate scenarios, critique assumptions. Iterate as needed, but stay anchored in the decision architecture.

You don’t have to build a system from scratch. Start by adapting a generic report format—just shift the sections toward your specific decision. Add a table, drop a constraint, sharpen a trade-off, then turn it into a prompt. The more clearly you define what the decision needs, the more precisely the AI can scaffold it for you.

This is how prompting becomes strategic design.

The Decision Divide

As decision scaffolds improve, the unit of meaningful work begins to shift. Not tasks. Not documents. Decisions.

Some roles are defined by this shift: choosing between futures, evaluating trade-offs, absorbing complexity into clarity. Others revolve around executing well-known tasks. The former will gain importance, the latter will face automation.

Look at your past week: how many meetings were about navigating uncertainty versus reviewing status? That split maps to the emerging divide. What if your week was measured by how many good decisions you made?

AI won't just speed up thinking. It will force us to decide what kinds of thinking still matter.