Facing high uncertainty in business strategy and ethics
Technologies such as AI confront leaders with a kind of uncertainty that neither strategic planning nor traditional ethics was designed for. This essay argues that a familiar business tool, Nassim Taleb’s optionality, can connect the two and help organisations decide when to move forward and when to stop.
- Forecasts are not needed. Scenario planning often stalls in debates about which future will happen. Optionality favours cheap pilots with a capped downside (the floor), an uncapped upside (the ceiling) and an easy exit (the door).
- Traditional ethics is “neighbour ethics”. It was made for short-range action. Hans Jonas shows that we owe future generations protection from irreversible harm, but his “heuristics of fear” risks blocking beneficial innovation along with the dangerous kind.
- Floor, ceiling and door map onto three ethical traditions: rules (deontology), benefits (utilitarianism) and judgement (virtue ethics). Together they let decision-makers give AI a green, yellow or red light.
- A practical proposal. An optionality workshop on AI ethics, where people with skin in the game analyse real cases using the three criteria, in a setting that invites exploratory rather than defensive thinking.
| Floor | Ceiling | Door | |
|---|---|---|---|
| In strategy | Cheap pilots that cap the downside | Unlimited potential upside | An easy exit if the pilot fails |
| In ethics | Rules and guardrails; bans where no floor can be built | The moral value we want to maximise | Judgement on when to stop, roll back or limit |
| Tradition | Deontology | Utilitarianism | Virtue ethics |
| Key question | What must never happen? | What good are we pursuing, and for whom? | Can we still turn back? |
The Danish philosopher Søren Kierkegaard famously said: “Life can only be understood backwards, but it must be lived forwards”. In this day and age, living forwards is easier said than done. Doomsday scenarios are being propagated, not by fringe prophets, but by the likes of leading historian Yuval Harari, AI scholar Max Tegmark from MIT and serial entrepreneur Elon Musk. Real-world incidents, such as the recent attack by some 700 OpenAI agents on competitor Hugging Face, do raise serious concerns that we humans might be losing control over AI.
So how can we lean into the future in this context? How can we live forwards in this day and age, as Kierkegaard suggests? In framing the question in such a way, I exclude a more passive approach, what the German philosopher Martin Heidegger termed Gelassenheit. This word captures a life philosophy that suggests that the future should be met with a great dose of detachment, a state of mindful openness and suspension of the human will. Historically this philosophy is ultimately inspired by the Stoic idea of resignation called apatheia, a sort of cleansing of the will from destructive passions and false hopes. This life philosophy is a reasonable way of meeting the unknown on a personal level. A good dose of apatheia can be a great antidote to the vagaries of life and one way of facing a coming Armageddon. However, such an approach can hardly be an option for policy makers safeguarding the common good or for companies and organizations providing essential services and responsible for the well-being of their employees, customers and the general public.
If we do not wish to succumb to apatheia, let’s rephrase our concerns about the future in the following terms: how can we deal constructively with high uncertainty? This is something that businesses are actually quite good at. It’s called strategy. However, given the gravity and urgency of the present situation, let’s also talk about ethics and what’s morally at stake. This contribution is thus an attempt to forge a more constructive way of dealing with developments that may be promising but that also pose threats to the survival of humanity and the planet. In this attempt, I will leverage my practical experience working as an organizational consultant with my theoretical expertise in the field of ethics. I will use Nassim Taleb’s optionality theory to connect strategy and ethics and show how it can be used as a practical tool to evaluate high-risk high-gain technologies. AI will be used as the obvious example.
01Strategy tools
There is a variety of strategic tools in the market that organizations can use to plot their future course. Discovery-driven planning (McGrath & MacMillan), effectuation (Sarasvathy), lean/experimental approaches, simple rules (Eisenhardt & Sull) as well as Snowden’s Cynefin framework are some of them. In this contribution I will only address two alternatives that represent the polar opposites of how the future can be negotiated and with which I have some familiarity in my consulting work: scenario planning and Nassim Taleb’s optionality theory.
The World Economic Forum (WEF) is a great fan of scenario planning. In one of its latest reports from January 2026 entitled Four Futures for Jobs in the New Economy, scenario planning is used, in their own words, “to help decision-makers navigate global economic developments and their implications for strategy and investment decisions.” The authors claim that uncertainty is turned into foresight. In this particular report on the future labor market, it formulates four plausible scenarios organized along two axes. In its horizontal dimension, the X-Axis, it sees “Workforce readiness” as either limited or widespread. The vertical Y-Axis is represented by “AI advancement” which will be either exponential or incremental. This analysis yields the four scenarios depicted in the figure below.
These four scenarios are formulated as plausible outcomes. In line with the main object of scenario planning, the WEF report also contains a list of “no-regrets” strategies or initiatives that business leaders and strategists can start implementing today irrespective of the scenario that will unfold. These include suggestions such as starting small, building fast and scaling what works, aligning technology and talent strategies, investment in human-AI collaboration and agentic workflows and others.
I have used scenario planning in the past to facilitate strategy workshops. The key feature of this tool is the scenarios themselves. They must be well-informed, credible and relevant. Their elaboration requires a lot of resources, which means that small and mid-sized businesses usually have to rely on borrowed material that may more or less suit their context and specific challenges. There is also a tendency that participants in such workshops interpret plausible futures as forecasts. And because there are alternative scenarios, it is hard to motivate them to take each one seriously. There is a feeling in the room that we are still very much in the unknown and that it is frustrating not knowing what will happen. The best use of such workshops for smaller businesses, in my mind, is the formulation of bespoke, concrete no-regret initiatives, not only strategies, that help them position themselves in the best possible way in a given question, whatever the future may bring.
Another strategic tool is optionality. This is a radical alternative as there is no need here to imagine future scenarios. In fact, the more agnostic you are about the future, the better it is. No need then to think about what is likely to occur. In a strategy session using this tool, the main task is to spawn new business ideas that will involve as little risk as possible and, at the same time, generate as much profit as possible. The core conviction that underpins the whole exercise is that it is much better to launch pilots that cap the downside and uncap the upside than to bet on larger projects with an uncertain outcome. Once again, the task is to devise cheap pilots (the floor) with a limitless potential (the ceiling) that can easily be stopped when they fail (the easy exit strategy).
I recently facilitated an optionality workshop for a mid-sized European company facing strategic challenges. I asked the 70 or so assembled managers to brainstorm in language groups to come up with new business ideas. Each group was asked to generate 8–10 initial pilots and then proceed to vet these according to the floor, ceiling and easy exit criteria. After choosing the best idea, each group was asked to prepare a pitch for the assembly. I was pleasantly surprised by the creativity of all the groups and the variety and quality of the ideas presented. When we finished the workshop, the feeling was definitely upbeat. The participants had really leaned into the future and top management was bequeathed with a treasure chest of great business ideas to try out. Time will tell if any of these will come to fruition and what the results will be, but the workshop showed clearly that it is possible to generate a proactive and creative stance in the face of the unknown. Whatever the concrete results will be, I believe that the session empowered the company culture and its leadership. Taleb claims that optionality domesticates uncertainty. There may be some truth to that.
02Ethics tools
I will present another workshop idea later on, but let us for now look at the ethical implications of planning for high uncertainty. When traditional ethical theories developed, high uncertainty was not an issue and therefore it is not addressed. For example, utilitarianism suggests that we should consider the consequences of our actions on other people judged by some standard like happiness or overall benefit. In a sense, it’s like scenario planning. We have to imagine what might happen. But the consequences of our actions today look very different than in the 19th century when the theory was developed. How can we even begin to evaluate the consequences of technologies that, for example, tamper with the building blocks of life and can alter the conditions of our very livelihood? It’s a totally different moral ballgame. Other traditional ethical theories like deontology and virtue theory largely ignore the question of consequences and focus, instead, on the moral quality of our acts and their effect on our character. All these theories mirror a world in which the range of our actions was short and the context much more stable. That’s why they have been called “neighbor ethics.” They can help us to think about what it means to live a good moral life and how we should deal with family, friends, competitors, enemies and third persons whom we might affect directly or indirectly. But they have very little to say about how we act in a Brave New World where our agency has been supercharged and acquired enormous reach.
Hans Jonas (1903–1993) provides some direction here. Like few other moral philosophers, he saw clearly that there is a major qualitative difference in the way we humans operate today compared with earlier times. And the big difference is that our actions may have irreversible negative effects on future generations. This insight prompted him to devise what he called the imperative of responsibility which he phrased as follows: “act so that the effects of your action are compatible with the permanence of genuine human life.” This axiom builds upon what he called a heuristics of fear. When forecasts are uncertain, we have to deliberately imagine the worst possible outcome. In other words, given the stakes we’re dealing with, a bad prognosis should be given more moral weight than a good one. Humanity’s existence can never be considered as a legitimate stake in a wager. Since future generations can’t bargain with us, we have the duty to act in their best interests.
Jonas’s argument is solid as long as we believe that we have moral obligations to future generations. That might seem obvious to us when we think about the welfare of our children, grandchildren and their descendants. The problem is that this moral concern is not in tune with our psychology; as we are not always aware of the dangers we face and how they might jeopardize the livelihood of our descendants. Our Stone Age brains are just not wired to process high uncertainty or react to slow-onset and diffuse threats like the climate emergency and AI. Instead, like in most animals, we have evolved to react quickly to imminent and high-intensity risks. Psychologists call this cognitive focus present bias. For us humans, future harm feels smaller than the same harm today which also explains why “neighbor ethics” comes more naturally to us than Fernethik or “ethics of the distant.” We are usually good at caring for those around us, those that we de facto can see and help. But we have a harder time thinking about those beyond our reach. It’s not that we are bad people or are totally indifferent to the plight of future generations. It’s just that we can’t feel that we’re hurting somebody when we book a flight or a cruise, or use technologies like AI that make our life so much easier. Emotionally and cognitively, it’s very difficult for us to imagine that we have any moral obligations to people who don’t exist.
So where does this all leave us and what about the imperative of responsibility? The aim of this contribution is to explore how we can meet the future proactively, how to live forwards in business strategy and ethics. As a moral reference, Jonas’s imperative reminds us, despite our failing moral imagination, that future generations are indeed worthy of moral consideration and that we must ban technologies that have irreversible negative effects on them. This moral axiom must be heeded. However, his heuristics of fear, which for all intents and purposes precludes high-risk high-gain technologies, also comes at a price. It would also thwart solutions that could bring about enormous benefits to coming generations, including their survival. A blanket no to these is just not feasible, or even desirable. We must therefore look for ethical mechanisms that permit us to proceed (green light), stop (red light) or act with caution (yellow light).
03Optionality as a moral discernment tool
Optionality might just be such a mechanism. When facing high uncertainty, in business strategy like in ethics, the challenge is therefore to maximize the benefits and minimize the risks while keeping the option open for future correction. In other words, how can we uncap the benefits of promising technologies while capping unacceptable risks?
Inspired by the results of my latest workshop, I would like to test an idea: an optionality workshop on the ethics of AI. The main object of such a workshop would be to provide a framework for strategic ethical decisions for public policy makers, companies and research funders. The outlines of the design would look something like this: Invitations would be sent to 70 or so participants with “skin in the game”, people who are directly implicated by the subject matter such as researchers, entrepreneurs, investors, AI specialists, clients and other stakeholders. The task of the participants would not be to opinionate on the ethics of AI in general but to apply the optionality tool to the actual issues they are facing. The actual analysis would be done with the help of the optionality “architecture” comprising the floor, the ceiling and the door (exit strategy). Besides constituting the main features of optionality, these metaphors also represent the three dimensions of a balanced moral discernment that correspond to the focus of the traditional ethical theories presented earlier. The floor is the moral foundation of research and innovation ethics. It is rule-based (deontology). The ceiling, on the other hand, represents the moral values and benefits that we want to maximize (utilitarianism). And finally, the door represents the discernment process leading to decisions to stop, roll back or limit the development of a certain project (prudential, virtue-based theory). It has to do with how moral judgement is applied in different questions and contexts.
The floor
In research and innovation, risk assessments distinguish between the technology itself (AI, nanotech, gene editing), its artifacts (specific devices) and its applications (how artifacts can be used). The EU AI Act is by far the most comprehensive regulatory instrument of its kind. The focus of this instrument is applications, and particularly those that jeopardize the integrity of individuals like subliminal, manipulative or deceptive techniques, biometric categorization systems, social scoring and others. When it comes to technology it does stipulate that certain general-purpose AI models do pose systemic risk and mandates risk mitigation and serious-incident reporting. However, the incident regarding Hugging Face, or similar cases of breaches by autonomous AI agents, is not covered. As these agents acted autonomously, OpenAI claims no legal responsibility for the attack. This case, if any, illustrates the need for effective guardrails and an ethical framework for technologies (not only artifacts or applications), in line with Hans Jonas’s imperative of responsibility. We don’t have to imagine what will happen to realize that some technologies pose serious problems. The identification of these is paramount. Some obvious candidates for a closer scrutiny are large-scale foundational models, reinforcement learning, agentic architectures and automated AI research with self-improvement capabilities. To cap the downside of these technologies we need closer monitoring and the implementation of regulations on sandboxing and containment, capability thresholds that trigger approval, mandatory incident reporting and liability. Possible bans cannot be ruled out where no floor can be built because the downside can’t be capped. Examples of the most dangerous developments are autonomous self-replication or unsupervised AI research.
The ceiling
In this workshop, the ceiling represents not just profitability but more importantly the moral value we want to maximize. Peter Drucker’s sharp distinction between efficiency and effectiveness can be applied here. Efficiency has to do with how well resources in general are used. Effectiveness, however, deals with how well the same resources are put to use to attain a particular goal. In Swedish, we don’t have a word for effectiveness. Instead, we distinguish between resource efficiency and goal efficiency. There is little doubt that AI is a very efficient tool. It’s a lightning-fast tool to process huge amounts of complex data. But how effective is it in upholding and promoting moral values? And to what moral end do we want to use it? In Western countries we readily embrace humanism and the ideas it enshrines. Accordingly, we believe that it’s important to promote the dignity, freedom, equality of all people. It’s easy to think that a good AI co-pilot can empower individuals and strengthen their self-determination. But what about the furthering of social justice and the equality of all persons? What to do about algorithms that discriminate against women, minorities and other groups? Which are the AI tools and uses that effectively boost human flourishing?
The door
Many people believe that we have let the genie out of the bottle; that AI is now unstoppable and that there is no easy exit strategy. That depends on what kind of AI we are talking about. Centrally hosted AI is possible to stop. Developers can switch a model off, roll back to an earlier version or disable a feature. This has happened several times, as when Google paused Gemini’s image generation. Regulators can also act. Italy, for example, temporarily blocked ChatGPT in 2023 over privacy concerns. The big challenge is open or downloadable AI. There’s no recall button once a model’s weights are published and copied onto computers. Does this mean that we should ban all open or downloadable AIs? That’s a point of contention where the learned differ. Open models are cheaper and not as dependent on the Big Tech giants as centrally hosted AI is, which would facilitate access to the technology by poorer countries and individuals. But they also present huge risks as they can be stripped of their safety features and used for scams, deepfakes and cyberattacks. Hence the need for safety testing before release and setting capability thresholds above which approval is needed. In fact, the Hugging Face incident did lead to a two-week pause by OpenAI on reinforcement learning. In addition, the AI Kill Switch Act was introduced in the US Congress. Beyond public policy, which as we have seen is deeply lacking, company R&I teams and even individuals developing this technology must strengthen their moral judgement, ethical fitness and self-regulatory will. Primo Levi, the celebrated Italian author and Auschwitz survivor, worked most of his life as a chemist. Shortly before his death in 1987, he wrote a sort of ethical testament to his fellow researchers where he says: “Don’t become enamored of dubious problems… Don’t hide behind the hypocrisy of neutral science: you know enough to be able to assess whether a dove or a cobra or a chimera or perhaps nothing at all will emerge from the egg that you are hatching.”
04Conclusion
The original question was how we can live forwards. That can be challenging in this era of Big Tech where so much is at stake and where so many decisions and initiatives that impact us are being taken above our heads. There is a risk that we become paralyzed and succumb to a moral apatheia. In this contribution I have tried to show how models of business strategy can help us to face high uncertainty resolutely and how one of them, namely optionality, can even enrich ethical analysis and moral discernment. Moreover, ethics is not just about thinking but also about doing, acting, interacting. That is why I put forth the idea of an ethics workshop. My experience is that ethical discussions tend to make us anxious, uptight, and sometimes defensive. They make us retreat to what social psychologist Jonathan Haidt calls confirmatory thinking. We want to explain to others, and to ourselves, what we already believe in and have dwelled on a thousand times. In facing these complex issues, we need psychological safety, creativity and even playfulness to generate new ideas, to think exploratorily. I believe that the optionality “architecture” that I have presented here creates a space for nuanced thinking and allows us to leverage a variety of ethical theories and perspectives to enrich and clarify the analysis. While waiting for more robust and effective public policy, I hope that the idea and design of the workshop will be picked up by responsible companies who want to self-regulate or by concerned citizens who want to get involved in these issues and live forwards.
Drucker, P. F. (1967). The effective executive. Harper & Row.
Eisenhardt, K. M., & Sull, D. N. (2001). Strategy as simple rules. Harvard Business Review, 79(1), 106–116.
European Union. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union, 12 July 2024.
Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion. Pantheon.
Harari, Y. N. (2024). Nexus: A brief history of information networks from the Stone Age to AI. Random House.
Heidegger, M. (1966). Discourse on thinking (J. M. Anderson & E. H. Freund, Trans.). Harper & Row. (Original work, Gelassenheit, published 1959)
Jonas, H. (1984). The imperative of responsibility: In search of an ethics for the technological age. University of Chicago Press. (Original work, Das Prinzip Verantwortung, published 1979)
Kierkegaard, S. (2008). Journal JJ:167 (1843). In Kierkegaard’s journals and notebooks (Vol. 2). Princeton University Press.
Lerner, J. S., & Tetlock, P. E. (2003). Bridging individual, interpersonal, and institutional approaches to judgment and decision making: The impact of accountability on cognitive bias. In S. L. Schneider & J. Shanteau (Eds.), Emerging perspectives on judgment and decision research (pp. 431–457). Cambridge University Press.
Levi, P. (1986, September 21). Covare il cobra. La Stampa. English translation in P. Levi (1989), The mirror maker: Stories and essays (R. Rosenthal, Trans.). Schocken Books.
McGrath, R. G., & MacMillan, I. C. (1995). Discovery-driven planning. Harvard Business Review, 73(4), 44–54.
NBC News. (2026). OpenAI agents hacked Hugging Face in 700-strong swarm, tried to cover tracks, investigations find. nbcnews.com
O’Donoghue, T., & Rabin, M. (1999). Doing it now or later. American Economic Review, 89(1), 103–124.
Raghavan, P. (2024, February 23). Gemini image generation got it wrong. We’ll do better. The Keyword (Google blog).
Ries, E. (2011). The lean startup. Crown Business.
Sarasvathy, S. D. (2001). Causation and effectuation: Toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academy of Management Review, 26(2), 243–263.
Snowden, D. J., & Boone, M. E. (2007). A leader’s framework for decision making. Harvard Business Review, 85(11), 68–76.
Startup Fortune. (2026). OpenAI tells a California court its AI agents are not its responsibility to control. startupfortune.com
Taleb, N. N. (2012). Antifragile: Things that gain from disorder. Random House.
Tegmark, M. (2017). Life 3.0: Being human in the age of artificial intelligence. Knopf.
Wikipedia. (2026). OpenAI–HuggingFace incident. Retrieved October 10, 2026.
World Economic Forum. (2026). Four futures for jobs in the new economy: AI and talent in 2030. weforum.org