RAY
We Are Very Good at Imagining Futures.
Now What?
We have scenario planners, foresight units, ethics boards, and science fiction writers on government retainer. Yet no one is in charge of the future.
A MAP OF FUTURES
Last autumn I stood in front of a speculative timeline at Goldsmiths CCA and felt unsettled. The exhibition was Life Before Automation by the London-based artist Lawrence Lek, running from September to December 2025. It brought together a decade of his films, simulations, and installations into a single immersive environment. What moved me the most was this commissioned piece at the entrance, spanning three centuries of AI development and merging real historical dates with imagined future ones: the gradual consolidation of AI into a fictional monopoly called Farsight, the incremental displacement of human labour, the quiet normalisation of machine decision-making across every domain of public life and a lot of the points that he had speculated to be plausible in 2045 had already come true by the time of the exhibition.
Image: Farsight is a speculative fictional timeline of AI by the artist Lawrence Lek, photographed by Pradipta Ray at the CCA exhibition titled ‘Life Before Automation’
Looking at this timeline posed a very persistent question for me: if this is a plausible trajectory for AI development, is it desirable? Should it be steered toward, or away from? What are the various technological tipping points, where these scenarios become more likely? These are not just abstract philosophical questions. They are exactly the governance questions that the UK’s emergent technology ecosystem is supposed to be equipped to answer, but after four months of interviewing futures practitioners, responsible tech practitioners, and ethics and risk practitioners of all sorts in the emergent tech sector in the UK, my research suggests it is not.
And this is the central paradox; the problem is not a lack of futures imagination. It is almost the opposite.
THE IMAGINATION TRAP
Two science fiction writers I interviewed in the course of this research both collaborate directly with government defence bodies, commissioned to imagine future conflict scenarios and stress-test strategic assumptions. Government departments run dedicated foresight units. Technology companies appoint responsible AI leads and convene ethics boards. Think tanks publish annual horizon scans and signals. Bioethics councils commission research on technologies that do not yet exist. Research institutes focus on civic participation so as to be inclusive and pluriversal in the futures we might imagine. Futures practitioners gather regularly to discuss long-term thinking and the politics of time. We are not short of imagined futures.
One might say that while we have a glut of futures work and thinking being produced, these are not new ideas. The renowned futures scholar Sohail Inayatullah coined the concept of the “used future”: an idea about tomorrow inherited without critical examination, resonant in its original context but no longer fit for the realities we are actually navigating. IDEO’s design futures team applies it to design practice
, identifying how organisations reach, almost automatically, for familiar images of what the future looks like - flying cars, glass-and-chrome smart cities, humanoid robots - without asking whether those images are still pointing at anything real. The futures they commission become, in effect, more sophisticated packaging for already-inherited assumptions.
The tools increasingly used to generate futures scenarios may be compounding this. Andrew Maynard, Professor of Advanced Technology Transitions at Arizona State University, coined the phrase “LinkedInification” in a March 2025 episode of the CIFS Farsight podcast to describe how large language models, trained on the vast corpus of human expression, regress to the mean: producing a standardised, professionally legible version of human identity scrubbed of eccentricity and strangeness. When these tools are used to generate futures scenarios, they risk producing visions that are similarly flattened, coherent enough to be credible, conventional enough to be comfortable, and conservative about what is actually possible.
While the debate about if anything is inherently new is one that can be had endlessly, the crux of the matter remains that the plurality of the futures we can imagine and whether or not they are stale is a moot one if we cannot think beyond just the creation of them. The deeper problem is that even genuinely rigorous and novel futures thinking - including Lek’s timeline and the many very thorough horizon scans being produced, are not connecting to governance. And this requires a different diagnosis.
FROM LONG-TERM THINKING TO GOVERNANCE
A distinction that is consistently blurred is the one between three different capabilities. The first is long-term thinking: for the purposes of this research, any organisation with a planning horizon beyond the next financial year is doing this. It is necessary but it is not, in itself, futures work. The second is what the futures scholar Riel Miller calls futures literacy: the capacity to hold multiple possible futures in mind simultaneously without collapsing into prediction. It is the ability to ask not, “what will happen” but “what could happen, and what does that mean for how we act now?”
The third is what governance researchers, following the framework developed by Ahern, call anticipatory governance capacity: the institutional infrastructure that makes futures thinking consequential. Regulatory frameworks designed to evolve alongside technologies. Governance functions built to experiment rather than simply observe and advise. Accountability mechanisms that connect what organisations imagine and commit to with what they are actually obligated to do.
A country, a department, or a company can be fluent in futures literacy and still lack any anticipatory governance capacity. The UK has made the repeated mistake of assuming that investing in the second would automatically produce the third. A government department can commission a rigorous horizon scan and file it without consequence. A technology company can appoint a responsible AI lead with no decision-making authority. A bioethics council can publish careful foresight research and watch it cited in ministerial speeches that precede no policy change. The imagination circulates, but the obligation does not materialise. The question of - ‘is this a desirable future, or one to steer away from?’ - goes unanswered not because nobody is asking it, but because nobody is structurally positioned to act on the answer.
THE SYSTEM AT WORK
Over the first four months of 2026, I spoke with fourteen eminent practitioners working across the UK’s futures and AI governance landscape, including government foresight practitioners, responsible AI leads at major technology companies, policy researchers at civil society organisations, and academics who have spent careers studying how responsibility gets embedded in large research programmes. What emerged was a set of patterns so consistent across sectors that they are difficult to attribute to individual failure. They look, instead, like features of a system.
In government, futures functions rise and fall with ministerial appetite rather than institutional need. The same work is commissioned, defunded, and rediscovered across political cycles, with expertise walking out the door each time. One practitioner described the situation with the kind of weariness that comes from watching the loop repeat:
“Futures in government goes up and down — really important, everyone is doing it; really unimportant, no one does it. They just spun the whole futures team of *** last year. And they were like, the leading light of futures thinking in government. Except, well, the new minister doesn’t like it.”
— Senior government futures practitioner
Futures outputs that do get produced rarely enter decision-making pipelines. Reports are circulated, noted, and filed. One practitioner described the structural problem precisely: there is a problem definition phase, a consultation phase, and then a report. There is no prototyping phase. The output is the endpoint rather than the beginning of a consequential process. The field is openly aware of this. Another practitioner, working across strategy and foresight, put the self-critique most bluntly:
“The discipline struggles with itself. How do I sell it? And we are in the business of creating narrative, but somehow we’ve trapped ourselves in the narrative.”
— Independent futures strategy practitioner who alos contracts with the government on futures projects
VOLUNTARY BY DESIGN
Each of these patterns has the same root cause. The UK’s responsible innovation and futures governance ecosystem is almost entirely voluntary. There are no mandatory requirements connecting futures work to consequential decisions. The relationship between anticipation and action is left to goodwill, organisational culture, and the taste of whoever happens to be in power.
“All of assurance is voluntary. None of it is mandatory. And that is also a key difference here.”
— Responsible AI lead, major UK technology trade body
The absence of mandatory accountability does not just create a gap. It actively shapes behaviour. The academic term for what results is performativity: the systematic divergence between what organisations say they are doing and what they are actually embedding. When nothing is required, the rational response is to do enough to signal commitment without building the infrastructure that genuine commitment would need. One researcher who spent years inside a major European publicly funded research programme described it directly: responsible research and innovation was used as a boundary object, because everyone had a different understanding of what it was. So they were just pretending.
Responsible innovation capacity in most organisations is also personal rather than institutional. It lives in specific people, not in structures. When those people leave, the capacity leaves with them. When financial pressure rises, the futures work disappears. The accountability for imagining and building desirable futures is not held anywhere that would survive a change of minister, a restructure, or a bad quarter.
NOT ALL SECTORS ARE EQUAL
The unevenness of this problem across sectors is one of the most revealing things the research surfaces. The financial services sector’s relative maturity in AI governance is not a cultural accident. It exists because mandatory accountability infrastructure was there before AI arrived: fiduciary duties, risk management frameworks, clear regulatory authority with enforcement teeth. These gave futures thinking somewhere consequential to go the moment it produced something useful. One responsible AI lead put the comparison precisely: financial services already has a risk management model, three lines of defence, and they simply apply it to AI. That is why the sector is more mature. Regulation is not an obstacle to responsible innovation. It is the scaffold that makes responsible innovation real.
AI and emerging technology sits at the other end of the spectrum. As one policy lead told me: “There is no existing AI legislation in the UK that provides an overarching framework. So, I would argue we don’t have the accountability mechanisms.” The futures thinking in this space is impressive but the governance infrastructure to receive it barely exists. A one-question diagnostic emerged from the research: ask who the responsible AI practitioner reports to. Data and legal teams signal a compliance function. The C-suite or board signals genuine capacity. The reporting line is a faster proxy for an organisation’s real relationship to responsibility than any published strategy. Whether something as simple as a question about a reporting line can meaningfully shift organisations is, of course, another matter. But it at least tells you what you are working with. Whether iterative anticipatory governance - regulatory frameworks designed to evolve in real time alongside the technologies they govern - can cover the gap between sectors as different as financial services and AI is the live question.
URGENCY REQUIRES AGILITY
The urgency here is real, because the technologies are here and progressing at speed that far outstrips any regulation. On the CIFS Farsight podcast in March 2025, Andrew Maynard described technological tipping points as phase changes between fundamentally different pathways to the future: moments when something introduced into a complex, interconnected system restructures it in ways that are difficult to reverse. He is not talking about AGI as an abstract horizon event. He is talking about what is already here. Over the last three years, he observed, we have seen a complete restructuring of society around how we use AI - if you look at things three or four years ago versus now, it looks very different. We are inside a phase change. The question is whether governance is ahead of it or behind it.
“It’s good to be able to anticipate when tipping points might occur. But it’s more important to develop an ability to be not only resilient, but also agile. The more agile we are, the more adaptable we will be as things change.”
— Andrew Maynard, CIFS Farsight podcast, March 2025
The Financial Conduct Authority’s approach to smart data regulation is what institutional agility looks like in practice. Rather than waiting for the technology to mature and then regulating it after the fact, the standard approach that has produced the accountability gaps visible across AI governance today, the FCA worked with innovators inside a regulatory sandbox, learning alongside them in real time. The outcome was that the Department for Business and Trade could draft the Data Use and Access Bill with knowledge generated through the process itself. Futures thinking had a customer. It had a decision-making pathway. As one practitioner put it: “Instead of letting the tech happen and then regulating, it works best when you do it together.” The sandbox is not a universal answer. But it is proof that the gap between imagination and governance can be closed when the institutional design is right.
THE REAL QUESTION
Returning to Lek’s timeline: nearly three centuries of AI development, plotted across a wall, asking which of these futures is desirable and which should be steered away from. The question it forces is not whether we can imagine our way to a better outcome. The imagination is not the constraint. The practitioners I spoke to across fourteen interviews - including the two science fiction writers working with defence, the government foresight leads watching their teams dissolved when ministers change, the responsible AI practitioners reporting to legal departments - are all, in their different ways, doing genuinely serious work. The constraint is that none of it is connected to a governance infrastructure that would make their imagination binding.
“We have a really good idea of what accountability mechanisms exist because we see them in other domains. But they currently do not exist in AI and emerging tech.”
— Policy lead, major AI research institute
Two trajectories are visible. In the first, the cycle continues. Futures functions rise and fall with ministerial appetite. Responsible AI leads report to legal. Voluntary frameworks multiply without producing accountability. The EU AI Act, with its mandatory requirements for high-risk AI systems, represents the alternative pressure: external accountability requirements that begin to recalibrate what UK organisations must actually do. Co-design models like the FCA’s accumulate evidence and get replicated. The reporting-line question becomes a governance standard rather than a practitioner heuristic.
Maynard argues that we do have the intellectual ability to nudge things preferentially in one direction or another. Central to that, he says, is an incredibly precious part of being human: the capacity to imagine different futures and then chart a pathway toward the ones we actually want. Futures thinking, he says, is exactly where that comes in. Maynard's point about tipping points is ultimately a modest one: we cannot predict which phase changes are coming, or how large they will be, or when exactly they will arrive. What we can do is build institutions agile enough to respond when they do. That is not a counsel of despair. It is the only honest account of what governance can actually achieve in conditions of genuine uncertainty.
Lek's timeline does not tell you what to feel. It just maps one plausible trajectory and holds it there, long enough for discomfort to set in. The futures plotted on that wall are not inevitable. They are the product of decisions being made, or not made, right now. The AI Safety Institute became the AI Security Institute. The responsible AI lead reports to legal. The futures team dissolved when the minister changed. These are not failures of imagination. They are decisions. The question is whether the next set of decisions will be any different, and who, structurally, is positioned to make them so.