1. Don’t Confuse Prediction With Knowledge
An AI prediction can look remarkably like a fact. It may arrive with a percentage, a chart, or an authoritative answer generated in seconds. But it is still an assertion about something that hasn’t happened.
“We have this idea that predictions are a source of knowledge,” Carissa says, “and we’re not taking into account enough how they are also a weapon of power.”
Predictions often carry an implicit instruction. A forecast that a market will collapse tells leaders to retreat. A prediction that an employee will underperform may influence whether that person is hired, promoted, or trusted with meaningful work. The output doesn’t merely describe a possible future. It encourages us to act as though that future is already settled.
Before following an AI prediction, ask what it actually tells you. Is it verified knowledge about the present? A probability based on past patterns? Or one possible interpretation of what might happen next?
2. Examine the Data—and What It Leaves Out
AI systems can process enormous quantities of data. That doesn’t mean they have all the information that matters.
“Predictions may be based on data,” Carissa says, “but even then you have to ask yourself what kind of data, and who collected the data and why, and what data might be relevant but we’re not collecting—maybe because it doesn’t exist, or because we don’t really want to look at it.”
Every dataset reflects choices. Someone decided what to measure, how to measure it, and what to exclude. The result may be extensive without being complete.
This is especially important when AI is used to make decisions about people. A system might evaluate an applicant using past hiring data, but that data could reflect years of organizational bias. It might predict a customer’s behavior without accounting for a sudden shift in their circumstances. It might produce a clean numerical answer while overlooking the context that a human decision-maker would immediately recognize.
Before acting, ask:
- Where did the data come from?
- What was it originally collected for?
- Whose experiences are well represented?
- What relevant information is absent?
3. Stay Close to the Present
The further a prediction reaches into the future, the more opportunities there are for the world to change.
“Try to tether your mind to the present,” Carissa says. “It’s very hard, but it often will give you a competitive advantage over others who are only thinking about the future—it’s much likelier that you’ll be good at making predictions close to the present than if you predict a thousand years into the future.”
Long-range AI forecasts can be seductive because they offer a sense of certainty. But predictions become more fragile as the timeline expands. An unexpected technology, political crisis, cultural shift, or new competitor can quickly make a confident forecast irrelevant.
Leaders also risk becoming so captivated by the technology they expect to have tomorrow that they overlook the technology they have today. Carissa points out that we often respond to flaws in existing systems with promises that those flaws will eventually be fixed. In the process, an imagined future can distract us from present performance, present risks, and present opportunities. So use AI to clarify what is happening now, and make near-term decisions you can review and revise.
4. Lead with Human Expertise
“Intuition is often a product of expertise,” Carissa says. “When you’ve spent decades studying something and you’ve seen many cases, you develop a kind of intuition that takes the data into account, but also has an understanding of the elements that make something come together in a way that an AI won’t have.”
An experienced leader, doctor, editor, investor, or teacher doesn’t rely on intuition instead of evidence. Their intuition is built from years of encountering evidence in context. They recognize quality, understand causation, notice anomalies, and know when a situation doesn’t fit the usual pattern. AI can help surface information, but it should not become a convenient way to avoid responsibility.
“Too often we externalize decisions on data or AI or companies,” Carissa says. “And while that might seem like a safer bet, you’re actually just surrendering your own good judgment and your own ability to be a leader. You’re being a follower.”
5. Prepare, Don’t Predict
The most consequential events are often the hardest to forecast. That makes resilience more valuable than certainty. “Think about how to make your life robust,” Carissa says.
Rather than asking AI to identify the one future that will occur, leaders can use scenario planning to explore several plausible futures. They can ask what each scenario would require, which choices remain useful across multiple outcomes, and what capabilities would make the organization more adaptable.
“When you’re thinking about the future, don’t just try to figure out which is the ‘right future,’” Carissa says, “but decide which future you’d prefer and figure out how to get there.”
Prediction treats the future as something to discover. Preparation treats it as something we still have a role in creating.
Want more from Carissa?
Watch her Lavin Voices podcast episode below, and get in touch to book her to speak at your event!
