The Creative Work in Scenario Planning That Can’t Be Outsourced

Don’t Rush to Ask AI for Ideation

AI has moved into professional fields faster than most of them could work out how to judge its use, and it is already embedded in everyday work, including scenario planning.

In 2025, the World Economic Forum and the OECD surveyed 167 foresight practitioners across 55 countries and found that two-thirds already use AI in foresight. Among those users, the most common applications are trend analysis and clustering (69%), scenario development (63%) and horizon scanning (60%).

I use AI in my foresight research as well: build agentic routines to explore and organise large volumes of scanning material, and use tools like Claude Skills to maintain research records, challenge scenarios and surface contradictions I might have missed, often before the coffee kettle boils.

The development of AI has made the future feel more tangible, and made people more willing to talk about it. But it keeps bringing me back to one question:

If scenarios are a medium for generating insight, challenging assumptions and supporting strategy making, does outsourcing the development process to AI reduce the chances of reaching insight?

Scenario planning involves two modes of thinking.

The first is analysis and sensemaking: scanning signals, identifying change, and working out the critical uncertainties and assumptions.

The second is creative imagination and narrative construction: recombining those uncertainties into a narrative sharp enough to challenge what an organisation already believes, and to change the direction of its strategy.

AI genuinely performs well in parts of the first one, though hallucination and source reliability still need to be examined manually.

Now people are bringing AI into the second mode as well: helping teams brainstorm and build new connections between conflicting signals, assumptions and worldviews, establish causal relationships, and turn them into full scenario narratives.

The question is when AI enters, and in what role.

Generating a scenario draft with AI does lower the anxiety of writing and can speed up the process. However, the statistical inertia of LLMs, together with bias in their training data, means their default output can potentially fall back on common linguistic associations and ‘popular’ narratives. This does not mean AI cannot produce something novel. It means that when the first few steps of imagining are handed over, the frame AI offers may become an anchor for everything the team thinks afterwards.

Before a team has really begun to diverge, it has already shifted from imagining different futures to editing the future the machine provided.

What is more risky than creative fixation is the outsourcing of the learning process itself.

Scenario planning is not just about producing a final set of scenarios. It is also a participatory intervention. Many of the a-ha moments happen while reading, listening, observing, arguing and reinterpreting. (That’s also why I enjoy participating in and facilitating forums and workshops where people come together to discuss possible futures.)

People gather and lay out all kinds of signals and trends, and try to understand the conflicts of value, assumption and emotion sitting behind them. Different people read the same material differently, drawing on their own experience, and the scenarios are iterated through those conversations.

These processes are not just about getting to an answer. They give a team a space to share opinions and revisit its own assumptions, and build ownership of the strategic choices that follow. If interpretation and sensemaking are frequently outsourced to AI, a team may well end up with a comprehensive report, without necessarily gaining the learning and insight that the process itself would have produced.

Don’t rush to the answer

Using AI is becoming as natural as searching on Google or scrolling social media. As information gets easier to reach, people may be less comfortable staying in a state of ‘not knowing’, and more impatient for an answer.

But in foresight, uncertainty itself is the material we work with. That is precisely why resisting the urge to turn to AI for an answer too early may become an increasingly important capability for foresight practitioners in this era.

Doing scenario planning in the age of AI, I think we need to keep asking ourselves:

• If we did this work ourselves, what would we learn?

• Do we need efficiency and convergence right now, or friction and divergence?

Scenario planning should still be human-centred in the AI era. Let people observe, interpret and diverge first. Then bring AI in to widen the range of evidence, challenge the logic, and inspire more possibilities for human imagination and conversation.