In today’s volatile and interdependent world, uncertainty is not a temporary disruption. It is a defining feature. Strategic foresight has become a critical capability for navigating complexity, enabling decision-makers to anticipate change, surface emerging risks, and imagine alternative futures. At this inflection point, Large Language Models (LLMs) offer a profound augmentation to how we conduct foresight, but not a replacement. In this context, the Policy Foresight Unit of the European Parliamentary Research Service has recently published “Augmented foresight: The transformative power of generative AI for anticipatory governance ”.
Let’s be clear. AI cannot foresee the future. It does not reason with intent or understand human values. What it can do is dramatically increase our speed, scale, and scope in making sense of potential futures, if used judiciously and critically.
The cognitive extension of human futures thinking
Generative AI can be understood as a cognitive prosthetic, a tool that extends human perception, memory, and pattern recognition. LLMs excel at identifying latent patterns in unstructured data, surfacing weak signals, and generating coherent narratives based on probabilistic inference. These capacities are particularly useful in exploratory foresight, where practitioners seek to widen the aperture of plausible futures. That is why, at 4CF we have created 4CF Sprawlr - the next-generation AI-powered debate and ideation platform designed to transform the way we brainstorm, strategize, and make decisions, as well as to challenge the assumptions of participants.
Moreover, during horizon scanning, LLMs can process thousands of documents across domains in real-time, clustering insights, filtering noise, and flagging emergent issues. This ability to synthesize vast and diverse information ecosystems supports faster and more responsive foresight cycles. But raw data synthesis is not foresight. True anticipatory intelligence requires contextual judgment, ethical discernment, and interpretive framing, capacities that remain uniquely human.
Narrative construction and scenario intelligence
Scenario planning, a cornerstone of foresight, benefits from AI in powerful ways. GenAI can help identify key drivers, test interdependencies, and co-develop scenarios that are internally consistent and plausibly disruptive. When used in parallel (“AI swarms”), different models can triangulate and refine assumptions, improving narrative robustness.
From a cognitive science perspective, LLMs excel at narrative generation because human cognition is fundamentally narrative-driven. We make sense of complex information by constructing stories, and AI's narrative fluency can mirror and support this innate process. However, plausibility does not equal probability or desirability, and this is where human oversight is critical.
Moreover, GenAI can model second- and third-order effects, encouraging deeper exploration of cascading impacts. But it struggles with ambiguity and values-laden trade-offs, areas where human foresight is indispensable.
New frontiers
One of the most intriguing applications of GenAI in foresight is the use of generative agents, simulated personas capable of expressing realistic, contextually grounded responses. These tools can be embedded in scenario narratives to create dynamic, dialogical futures. Emerging research shows that synthetic respondents can replicate human survey responses with high fidelity in controlled environments. They enable simulations of behavioral dynamics in policymaking, urban planning, or public health where direct access to participants is limited. Yet, the epistemic status of these agents remains debated. Are they truly proxies for human complexity, or merely statistical echoes? Their outputs must be seen as exploratory artifacts, not empirical evidence, informing, but not determining, foresight conclusions.
Bias, blind spots, and the illusion of objectivity
Despite their capabilities, LLMs come with significant epistemological risks. Their training data is historically bounded and socioculturally skewed. Without critical intervention, they risk amplifying dominant worldviews and suppressing marginalized perspectives, exactly the opposite of what good foresight should do. This reinforces what foresight scholars have long argued: that the future is not a neutral space, but one shaped by power, values, and competing imaginaries. AI outputs can reflect and reproduce systemic biases unless de-biased through intentional prompt design, algorithmic transparency, and participatory governance frameworks.
Moreover, the fluency of LLM-generated text can create a false sense of authority, leading to overreliance or reduced critical engagement. Practitioners must remain reflexive and iterative, treating AI outputs as inputs for dialogue, not doctrine.
Augmented intelligence, not artificial intuition
There’s a philosophical tension at the heart of this conversation. Foresight is not merely an analytical process, it is also a normative, imaginative, and ethical act. It involves asking: What kind of future do we want? Whose futures are we considering? What trade-offs are we willing to make? Generative AI does not possess values. It does not dream, hope, or fear. These are deeply human faculties, and they are essential to meaningful foresight. That said, when used well, GenAI can be a strategic co-pilot: accelerating discovery, enabling richer scenario exploration, and expanding access to foresight methods across disciplines and organizations.
Co-evolving with the machine
In sum, the integration of GenAI into foresight practice should be viewed not as a technological leap, but as a sociotechnical evolution. The most effective foresight processes will be those that combine human insight, ethical reasoning, and narrative richness with the analytic power and generative capabilities of AI. To paraphrase a growing sentiment: the future will not be written by AI alone, but by humans who know how to work with it wisely. The real promise of GenAI lies in partnership, not replacement. Because while AI may help us see further, it is still up to us to choose the path, interpret the terrain, and navigate uncertainty with courage, care, and imagination.
