Designing for the 10-Year Question

"Where will we be in ten years?" is the question that breaks most strategy decks. Nobody can answer it. Everybody pretends to. The pretending is what costs.

Here's a framing that's worked for me, with apologies to the people who invented scenario planning long before I came along.

Don't predict. Identify forces.

Ten years out, you can't reliably predict which technology will dominate, which competitor will be ahead, or which regulation will exist. But you can usually identify the forces that will shape any version of the future, and which way each is likely to move.

For most strategic questions, three or four forces dominate. Examples in my domain:

• Compute cost (down a lot, down a little, flat)
• Open-source frontier-model parity (caught up, stayed behind)
• Regulatory burden in the EU (light, heavy)
• Talent supply (abundant, scarce)

Build the 2x2, but expect more

Pick the two forces with the most uncertainty and the most impact. Plot them as axes. You get four scenarios, one per quadrant. Each scenario is a coherent future, not a prediction.

The 2x2 is a starting point. Real foresight work usually ends up with 5–7 scenarios, including a "current trajectory" and a few low-probability shocks. The point is to make the team think across the space, not to land on the right point.

The "what would we do" question

For each scenario, ask one question: what would we do differently if we knew this was happening? If the answer is "nothing," you can probably ignore the scenario. If the answer is "everything," it's a scenario you need a contingency for.

This is where foresight becomes strategy. The point isn't to know which future is real. It's to know which futures would require a different posture, and to be ready to recognize the signals when they appear.

The signals layer

For each high-impact scenario, identify 3–5 leading indicators you can monitor. "If compute cost falls 30% in 18 months, we're heading toward Scenario A." That gives you a switch you can throw when the data starts pointing somewhere.

Without indicators, foresight is just creative writing.

The deliverable isn't a forecast. It's a watchlist.

Why this works

The framing works because it separates the parts of the question that are answerable (which forces matter, what counts as evidence) from the parts that aren't (which scenario will be real). You stop trying to be a prophet and start being a strategist with options.

The 10-year question never gets a clean answer. The team that's thought about it tends to do better than the team that hasn't.