In the world of strategic forecasting, there’s a recurring problem: the questions we’re asking today often evolve into the wrong questions tomorrow.
Developing a strong forecasting question portfolio has always been a labor-intensive, and therefore expensive, process. It typically involves several stages, each demanding significant expertise and time:
- Research and analysis to ensure we’re asking the right questions. If you’re trying to understand the trajectory of the US/Iran/Israel conflict, for example, what is the right set of questions to ask? Do you ask about the Strait of Hormuz? Attacks on civilian infrastructure? Something else? Do any of these questions even give you a clear sense of the future trajectory of the conflict? This frequently involves a structured issue decomposition process - breaking a broad strategic concern into its component drivers and indicators.
- Extensive refinement of each question to make it truly forecastable and resolvable. Frequently, people want to ask questions like “What will happen with the Strait of Hormuz?” – a question that begs for an answer, but is far too vague to forecast. We need to refine this into something forecastable without losing the underlying question intent.
- Resolution design, including identifying suitable resolution sources, defining unambiguous resolution criteria, and providing clear background information so forecasters have the context they need.
And that’s all before a single forecast is produced. Once questions are published, human forecasters invest their own significant time researching each question and generating well-calibrated probability estimates.
The Sunk Cost Trap
All of this time investment creates an underappreciated problem: your question portfolio can’t be very agile.
When the situation on the ground shifts - a new actor emerges, an unexpected escalation changes the dynamics, a diplomatic breakthrough reshapes the landscape - the ideal response would be to reshape your question portfolio. Ask different things. Focus on what matters now, not what mattered when the portfolio was designed weeks or months ago.
But you can’t just throw away questions that dozens of people spent hours developing and forecasting. It would be wasteful of all that time, effort, and expense. And even if you were willing to do it, you’d need to reinvest the same amount of time to develop and forecast a new set of questions - by which point the situation may have shifted again.
So instead, organizations keep forecasting on questions of diminishing relevance, because the cost of starting over is too high.
AI Changes the Calculus
This is one of the most underrated benefits of AI-powered forecasting tools - and it’s a capability we’ve built directly into our platform, Hinsley.
In just a few minutes, Hinsley can:
- Perform deep research on a topic to gather the latest available information from across the open source landscape.
- Generate a structured decomposition of the issue, identifying the key drivers, indicators, and uncertainties that matter most.
- Produce forecasting questions for those indicators, each with resolution criteria, background information, and appropriate time horizons.
- Begin producing regular forecasts on those questions using state-of-the-art AI forecasting models.
- Still solicit human input and oversight on the questions and forecasts, as appropriate.
What used to take weeks of analyst and forecaster time now takes minutes.
The Real Unlock: Permission to Start Over
This speed doesn’t just save time - it changes how you can approach forecasting.
In the past, we’d have significant heartburn about voiding questions in our forecasting tournaments, and rightly so. We’ve always wanted to be respectful of our forecasters’ time and effort. But with Hinsley, if the situation evolves and we believe a different set of questions would have greater discriminating power about a topic’s future trajectory, we can simply throw away the old questions and forecasts without guilt, wasted effort, or weeks-long lag while new questions are developed.
This is a real AI forecasting superpower: not just accurate forecasts, but the freedom to treat your question portfolio as a living, adaptive instrument - one that evolves as fast as the world it’s trying to anticipate.
The best forecast isn’t always the most precise one. Sometimes it’s the one that’s asking the right question at the right time. And now, for the first time, we can afford to make sure that’s always the case.