Brian Stauffer
The ability to engage with radical uncertainty is becoming increasingly urgent for leaders — and they need tools for doing this rapidly. Scenario planning is growing in popularity as a way to better understand and prepare for changing business contexts, but, as practiced in the past few decades, it can be too time- and resource-intensive to meet today’s need for fast insights.
While thoughtful attention and consideration of rapidly evolving issues is more important than ever for managers, dedicating people’s valuable time to produce a set of scenarios is becoming increasingly expensive. It is also ever more difficult because many managers are stretched thin — a situation that’s exacerbated by the delayering of central management and strategy teams.
How, then, can managers produce a bespoke and useful set of scenarios to address their own specific situations and uncertainties? And how can they do so without expending vast resources, managerial attention, and person-days — or waiting six months or more for useful insights?
A few companies have found a way to do all of the above. Their approaches involve framing scenarios with their intended users, those engaged in strategic planning, at the center, placing a sharp focus on what really matters here and now to those individuals. These companies have also made judicious use of generative AI tools in the process. We will explain how these elements are being applied in an accelerated scenario planning process, illustrated by the recent experiences of two companies: Fazer, a Nordic fast-moving consumer goods company; and Unum Ltd., the U.K. subsidiary of a U.S.-based provider of employee benefits.
Framing Scenarios for Their Users, Here and Now
As the discipline of scenario planning has evolved, so, too, has thinking about how to frame the purpose of a scenario: that is, what it should be about or who it should be for.1 A particular approach that has emerged is to tailor scenarios not to the organization in general or to a given generic topic but to the person or people whose decisions will be informed by those scenarios. This approach puts the unique worldview of the intended user or users at the center, because each person’s specific worldview underpins their perceptions and actions on behalf of the organization. The power of scenario planning in this context is to help those users surface, question, enrich, reframe, and reevaluate their worldviews.2
In this article, we build on a refinement to this approach, which is to focus on challenging the user’s ghost scenario.3 A ghost scenario is the existing, often implicit scenario underlying a company’s current investments or strategies, comprised of the assumptions senior people already have made about what the future environment is going to be like. It is these assumptions that they will want to examine if they undertake scenario planning. When someone focuses on their own specific assumptions, the work becomes anchored in their current mental map.4 Challenging and contrasting these assumptions with other plausible ideas makes the new scenarios immediately interesting and useful.
This much sharper, user-centric approach quickly focuses scenario planning teams on establishing which small set of assumptions in the existing ghost scenario users will find most interesting and salient to challenge in the here and now. These assumptions, surfaced in conversation with senior management, may relate to gut feelings that the company is over-reliant on a particular trend continuing, or they could represent wishful or outdated thinking.
In the case of Fazer, the scenario planners identified the ghost scenario by interviewing senior managers and making a list of 15 assumptions that they all agreed were held in the business. The planners then conferred with the management team (the main users) to narrow the list to three or four assumptions that they should focus on. In Unum’s case, on the other hand, a review of existing strategy documents, board minutes, and company announcements yielded a list of assumptions. After they were validated by senior management, the strategy director selected two for exploration, with the final approval of senior management.
Focusing on a very small number of core assumptions related to a specific purpose and building scenarios that directly challenge users’ existing ghost scenarios makes the work immediately relevant and useful to those making important decisions on company direction. It avoids scenarios that boil the ocean and prevents longer-term views from being unrelated to what matters now. This is important given that strategic failure is highly correlated with senior managers making choices based on incorrect assumptions.5
Speed Up Scenario Development With AI
Using AI tools based on large language models (LLMs) in developing scenarios is now the subject of much research.6 In the cases we studied, the tools were used to research factors that might either challenge or reinforce the assumptions in the ghost scenarios and to generate scenario outputs.
The output for each scenario developed is typically a narrative told from the perspective of someone in the future, recounting what happened that led to that future state. In the scenario planning methodology we have developed at the University of Oxford, a useful scenario is plausible, not probable.7 Creating scenarios is an exercise in stretching the imagination to consider what could happen, not predicting the likelihood of outcomes.
Research to support scenario planning typically happens after the core assumptions to be challenged in the ghost scenario have been identified, as well as during the development of the scenarios. (See “A Focused, AI-Enabled Scenario-Building Process.”) This research explores relevant technological, economic, political, social, environmental and legal factors that could unfold in the future, looking especially at developments that could challenge what is currently being assumed by executives. The research is aimed at sparking imaginative thinking and broadening people’s perspectives.
After scenario planning workshops have concluded, the highlights of presentations and other building blocks of the scenarios can be fed as prompts into a generative AI tool, which can produce a draft narrative. This may be done a few times to refine each scenario and to distinguish each from the others. In addition, the scenario narrative produced by GenAI can be fed into a video generation tool, making it easy to tell the story through a more accessible medium. We estimate that the time that would have been required to produce both of these outputs via manual effort was reduced by 50% at Fazer and 40% at Unum by instead using generative AI. Moreover, the AI models enabled additional elements to be incorporated into the work, such as video and image generation, which would not have been the case working without AI. In both scenarios, the refined outputs were judged to be of sufficient quality to be incorporated into reports and presented to senior management to support and inform strategic planning.
How Fazer Accelerated Scenario Planning
Exposure to several different markets and industry categories opens a company to many unpredictable developments. This is true for Fazer, a 134-year-old Finnish consumer goods company with a broad product portfolio that includes confectionery, bakery, and breakfast products; plant-based drinks; flours; and various food innovations.
The company sought to thoroughly assess plausible future scenarios to improve its readiness for situations that were not easily foreseeable. It also wanted to energize its annual strategy conversations with new perspectives that could support ambitious growth plans. Fazer decided to adopt the Oxford approach based on the ghost scenario because it held promise to generate meaningful scenarios without consuming an inordinate amount of resources and managerial time.
Initially, discussions meant to inform the strategy process investigated what-if situations like “double-digit volume drop,” “radical margin drops,” or “significantly weaker demand.” However, it soon became clear that those questions were too self-focused and didn’t truly stretch people’s strategic thinking.
The Oxford approach shifted the focus to the broader context the company might inhabit. The Fazer strategy team, which later led the scenario planning effort, considered assumptions that had remained unquestioned in the strategy, such as “What if one or more key markets changed dramatically?” or “Could the raw material upon which the portfolio depends shift materially?” Because the company produces a wide range of consumer packaged goods, many of its ghost scenario assumptions focused on the supply of inputs, primarily raw materials. These discussions helped the team outline what research was needed to support building plausible scenarios for such situations. Ahead of diving deeper into the scenario work, the ghost scenario and related aspects were discussed with senior management. By the end of March 2024, a working group consisting of 20 handpicked senior representatives from supply chain, procurement, R&D, marketing, finance, and communication convened for a two-day workshop, where eight of the 15 assumptions benefited from additional participant perspectives. The insights from the session were further analyzed and explored with support from LLM tools.
One month after the first workshop — that is, within two months of beginning the scenario planning process — the Fazer scenario team had created a half-dozen plausible futures that challenged key assumptions in the ghost scenario. (It subsequently eliminated a few that were deemed insufficiently unique or relevant to the core assumptions.)
The team went from exploring possible developments, such as rising sea levels, changing cultivation conditions, and rising temperatures, to shaping a comprehensive view of a future where climate change resulted in significant negative impacts on the crops on which Fazer’s businesses depend. The team developed a specific scenario on climate change that enabled it to explore how countermeasures for addressing such changes may fail or prove insufficient. That process reframed the Fazer team’s view of climate change as primarily an issue to be mitigated by targeting emissions reductions and made explicit what had been less in focus: the somewhat inconvenient truth of the drastic impacts of these complex changes for Fazer’s suppliers, its own operations, and its customers. This allowed the company to consider a comprehensive view of key impacts that would have otherwise gone unnoticed, including quantitative projections for future crop-growing regions.
Management teams at Fazer have been able to use the scenarios to explore alternative futures and how they see their work within them. They have questioned aspects of the strategy that had been taken for granted that were related to some of the most fundamental elements of the existing business model.
Strategic dialogue at Fazer involved active reflection against the new scenarios, with each business expected to outline key topics to be addressed across each of them. One key lesson was that the use of scenarios needs to be integrated into existing processes for strategic assessments. Otherwise, they are too easily put aside. But in this case, the scenarios generated compelling insights, enabling critical thinking to help executives challenge their underlying strategic assumptions, and generated ideas that differ significantly from current strategy.
How Unum Challenged Its Own Assumptions
At Unum Ltd., increasing uncertainty in the operating context prompted executives to challenge their assumptions as they developed the next strategic plan. The U.K. subsidiary of the U.S.-based insurer Unum Group, Unum Ltd. is a specialist provider of employee benefits, offering income protection and critical illness, life, and dental insurance to over 2.8 million people across the U.K.
Unum Ltd. had grown significantly following the COVID-19 pandemic. Capitalizing on a buoyant U.K. employment market and heightened corporate interest in employee health and well-being, the company aligned its value propositions with evolving customer needs and outperformed its larger, more diversified peers.
By the time Unum Ltd. was approaching the next strategy cycle, its operating context had shifted. It faced a sluggish U.K. economy, slowing employment growth, and mounting pressures on the public health care system. Added to this were uncertainties surrounding the adoption of AI and its implications for the U.K. employment market in the years to come.
Unum Ltd. launched its scenario planning exercise in January 2025 with the goals of exploring the key uncertainties shaping its operating context and testing whether its strategic choices would be resilient across a range of potential economic, political, and social futures. In order to meet a May deadline for delivering scenarios and a preliminary draft of the 2030 strategy to the board, the strategy team fast-tracked the development of a set of scenarios to be ready for mid-March so that they could effectively inform and stress-test the strategic choices under consideration. The team opted for the accelerated approach described in this article, focusing on key assumptions in the ghost scenario and using AI to significantly reduce the required time investment.
The initial step was to surface the ghost scenario. The strategy team extracted assumptions from a recent survey of (and interviews with) the executive team that had explored future trends. Executives’ perspectives largely reflected an extrapolation of current trends, with a moderate level of disruption to the employment market from AI. The underlying ghost scenario assumptions were grouped by topic, which helped to narrow further research to 15 focal topics ranging from future trends in U.K. health care provision to the future of employment in an AI-driven corporate sector.
A full-day scenario development workshop held in February brought together subject matter experts from across Unum Ltd., including sales, operations, marketing, and technology, as well as representatives from a public affairs consultancy and an executive with recent leadership experience at one of the company’s closest competitors.
Before the workshop, the strategy team summarized the 15 key focal topics previously identified, covering key political, economic, societal, and technological trends. To streamline the research process, it used LLMs to identify data sources for each topic and then engaged directly with those sources to assess their credibility and obtain firsthand insights.
During the workshop, the strategy team presented the key hypotheses for each focal topic, along with opposing trends that could challenge those assumptions. Participants were encouraged to critique the content and contribute their own perspectives, including any additional trends or insights they had based on their own knowledge, research, and personal experience.
Participants were then split into groups and given cards representing the opposing side of each hypothesis. For example, one card might state that “AI drives the creation of high-skilled jobs” while its counterpart contended that “AI contributes to widespread job displacement.” Their task was to map a sequence of events that could plausibly lead to a specific set of outcomes by 2040. Extending the timeline beyond the 2030 strategy horizon encouraged broader thinking and opened the door to a wider range of potential scenarios. Each group then presented its plausible logical maps to the wider workshop. Scenarios that shared overlapping themes or led to similar outcomes were consolidated into unified maps to capture the most representative futures.
The resulting scenarios tested two key assumptions: the extent to which AI could influence demand for human capital, based on whether AI will primarily displace existing jobs or create new, high-skilled job roles that contribute to higher productivity; and the level of government intervention likely to occur, with an emphasis on the labor market. For Unum Ltd., a provider of employee benefits that function alongside state-funded health care and social security benefits, these assumptions were identified as the most critical uncertainties shaping its future operating context.
Following the workshop, the strategy team refined the scenarios to ensure that each one sufficiently contrasted and challenged the ghost scenario. The final set of scenarios (consisting of the ghost scenario and three alternatives) were then fed into an LLM to craft vivid narratives depicting what life in the U.K. might look and feel like in 2040 under each scenario. These narratives captured both the broader contextual environment and the specific market dynamics relevant to Unum Ltd. The team also made three-minute AI-generated videos based on the narratives that illustrated life in the U.K. in 2040 under each scenario and described the key developments that had led to that scenario.
The strategy team presented the scenario set to the executive team in March 2025 so that the strategy under each future state could be validated and evaluated for potential risks and opportunities. The videos proved to be a powerful engagement tool, facilitating meaningful discussion and enabling the executive team to grasp the nuances of each scenario. In contrast to text-based scenario descriptions, the visual format reduced the risk of misinterpretation and helped quickly establish a shared understanding of the scenarios.
For Unum Ltd., the scenario planning exercise proved its value in two key areas. First, it provided greater clarity on how the most critical uncertainties in its operating context might influence the strategic decisions under review. And second, it demonstrated the potential of AI to accelerate complex strategic processes, reducing the number of process-driven tasks to enable a greater focus on higher-value strategic thinking.
Lessons Learned Using GenAI in Scenario Planning
We advise those using LLMs in this context to approach them not as “objective” sources of factual information for automating research but rather as collaborators in imagining the future in creative ways. That is, think in terms of “AI-in-the-loop” rather than “human-in-the loop.”8 This subtle difference puts humans at the center of reasoning processes, with LLMs helping in ways that are deemed useful in a specific situation. (Executive assumptions and strategic options are always unique to an individual organization.) An AI-in-the-loop perspective treats LLMs not so much as authoritative sources but as partners for brainstorming ideas about the future.9 Some ideas may be fanciful, but that can also be the case with ideas generated by humans alone. LLMs can provide executives with quick access to an ideation partner whose resources extend well beyond the company’s boundaries and the limits of groupthink.
Unum Ltd. used personas in its prompts to generate new perspectives: “You are an expert scenario planner looking to challenge the assumption that [ghost scenario assumption]. Provide a set of compelling arguments that support and contradict the assumptions, providing links to the underlying sources, such as research papers from credible experts and international institutions.” For Fazer, when developing the scenarios from the 200-300 word descriptions generated in the workshops, one prompt that proved helpful asked the LLM to “make this alternative future description more captivating based on insights X, Y, Z” and to “give a second opinion on the description of the scenario, providing support for the reasoning behind the description, as well as counterarguments as to why this future may not be plausible.” A key insight from both the Fazer and Unum cases is recognizing prompting as a running thread and not a one-off activity.
Evaluating the outputs of LLM queries then comes down to assessments of plausibility (in the context of relevance and usefulness to the case at hand). Scenario planning is conducted in situations of deep uncertainty, where judgments about what is probable in the future are impossible to make. Yet, human assessments can be affected by biases or a lack of knowledge, and organizations must find ways to counteract these issues. For example, in judging the LLM outputs of the scenario descriptions, Fazer identified and set up discussions with external experts to review the descriptions in order to validate the insights and underlying data. Unum Ltd. treated the LLM outputs as “raw” and reviewed the documents that the LLM had cited as credible sources to help the company assess the value of what was being proposed. The experiences of both Fazer and Unum attest to the importance of considering how best to evaluate LLM outputs, and of acknowledging that the users of those outputs may have limited knowledge or biases that result in ambiguous or problematic evaluations.
While scenario planning is more necessary than ever, given the increasing uncertainty in the business environment, demands on managers’ time and attention are growing quickly too, making it hard to commit to new practices. New approaches to scenario planning are needed to reconcile these two forces. We believe that the three elements we have outlined — bringing the longer-term view into the here and now; framing the scenarios by questioning key assumptions in the ghost scenario, thus rendering it explicit and questionable; and judiciously using AI to produce the scenarios more effectively and quickly — can help leaders focus their scenarios on the immediate needs of their intended users.