Should we trust synthetic advisors?
· 12 min read
This is article 1 of 4 in What the Future Wants from Us series.
How do we best integrate AI into stakeholder capitalism?
We have been here before. Market research already leans hard on what are variously called "synthetic respondents," "synthetic panels," or "AI personas"—pre-testing concepts, messaging, and pack designs before spending on fieldwork, simulating hard-to-reach segments, stress-testing survey instruments, and filling gaps in panel data. Tellingly, both the Market Research Society and NIQ (NielsenIQ) have published cautions. Skeptics have been blunter: one calls such panels "the homoeopathy of market research."
The question I have been chewing on is whether AI will tempt the sustainability field to try something similar. What follows is a notebook on some early—and very basic—trial runs by an AI novice.
Over the decades, I have served on more than 80 boards and advisory boards, which taught me what real-world brain trusts are for, how they work at their best, and what happens when they misfire. Believe me, some do.
To be clear, I am not advocating virtual advisory boards, least of all as a cure-all. But I suspect they are coming, like it or not, so we should be thinking through the implications and applications before others beat us to the draw.
If the sustainability agenda is about anything, it is about what the future would want from us—if it could vote for, invest in, or work alongside those creating better conditions for all life forms, including the unborn. Hence the title of this four-part series.
So where is the sustainability sector headed, and where should it be headed? I explore some answers as lead author of a forthcoming Green Paper, part of the 'Routes to Transformation' initiative our London-based change agency Volans is running with GlobeScan and the ERM Sustainability Institute. One aim is to stretch our collective focus out to 2037, rather than the more usual 2030—the year we will almost certainly be forced to concede that the UN's Sustainable Development Goals have fallen short.
More on that in the second post.
To pressure-test that bet, I wanted a panel designed to argue back. So, I built one, using two rival AIs: Anthropic's Claude and (holding my nose) OpenAI's ChatGPT. The result is a Virtual Youth Advisory Board. Six composite young people distilled from well-run global youth surveys, plus a rotating seventh chair for a more human perspective, even if that person too is a figment of my imagination.
The technology is evolving fast, whether we like it or not, with 4D video avatars now on offer. So it's time to think all of this through.
The early response from young colleagues was, unsurprisingly, negative. They saw a real risk that their generation could be disintermediated en masse, like data entry clerks, technical translators and taxi drivers.
So, before I introduce my six avatars, an awkward question deserves a straight answer—the one I would ask if someone else had come up with this. Should anyone trust an advisory board of people who do not exist?
Start with access. Nobody can convene six continents' worth of twenty-somethings for every question a research programme throws up. Real advisory boards, whether young or older, are expensive and slow to assemble—and precisely because they are expensive and slow, they tend to be consulted rarely and late, when the strategy is already largely written.
Then there is what the method does to disagreement. Surveys average people out: ask a thousand young people whether they are optimistic and you get a number that nobody actually holds. Put six sharply drawn positions in one room and the disagreements survive—and the disagreements are often the most useful output.
Third, it reaches constituencies no consultation currently reaches. No real-world board offers a seat to a river, a beaver, an ecosystem or a citizen of 2050. Ours can—and in at least one case does. It also iterates in minutes. I have run ten sessions in a few hours; no real board could.
Now the objections, put at full strength rather than as straw men.
Ventriloquism. The most serious charge. These may be my own views wearing six costumes, generated by systems trained to be agreeable, prompted by a white, 77-year-old man with a 50-year stake in the answer. The avatars may have been compiled from well-founded research worldwide—but that is not an answer.
Substitution. Once an organization has a synthetic panel on tap, does it still fund real ones? The method could license exactly the corner-cutting it claims to remedy: a board that never costs anything, never reschedules, and never says anything the chair can't live with.
Training-data bias. "Young people worldwide," distilled from surveys and the models' own heritage, over-weights the online, the English-fluent, the educated and the surveyed. Kwame is not Africa. Mei is not Asia. At best these are heuristics; at worst, caricatures with names attached.
No accountability. A composite cannot withdraw consent, be misquoted, sue, or be proved wrong. None of the six will ever say "you misrepresented me"—which removes a key check on anyone who claims to speak for others.
Consultation theatre. Youth activists already level this at real boards: flown in, workshopped, photographed, thanked, ignored. Synthetic panellists make it worse, because they cannot even walk out.
Now for the context. The Black-Swan-going-Green image below is my effort to capture the central dynamic of the coming transition. It will be uneven, falter badly at times, and be hotly contested wherever it threatens incumbents—and it most certainly will. But the trajectory is clear and increasingly exponential, toward breakdown and breakthrough at the same time.
What our 2026 survey results show is not so much a series of impending Black Swans as the high-probability, already-visible, grinding failure of the mainstream trajectory—emissions still rising, progress too slow, the agenda often mired in reporting.
But maybe the survey has a blind spot? What might the Black Swans of the next ten to fifteen years look like? The options are legion. A cyber-physical "Pearl Harbor," perhaps, or a cascading attack on grids or on undersea cables carrying intercontinental data. Or imagine a foundational digital system degrading beneath economies that depend on it—GPS, say, or the Cloud.
Then there might be the sudden loss of a pivotal leader, a genuine discontinuity now that so much of politics is personalized. Or a China–Taiwan blockade, rather than an invasion. Or the 2028 US presidential election tipping into sustained political violence; the next pandemic, maybe with H5N1 acquiring human-to-human transmission; a single tactical nuclear detonation, shattering an eighty-year taboo and resetting every risk premium; or a climate step-change, with compound failure across several breadbaskets at once.
More positively, a spectrum of Green Swans is surfacing too. The so-called "green premium" is collapsing category by category—solar, EVs, heat pumps—until the clean choice is simply the cheaper one, pulling storage, grid, heat and transport along in its wake.
On the demographic front, an estimated $80–140 trillion could pass to younger, more values-driven generations by 2045. AI is compressing materials discovery, grid optimization and climate modelling. Carbon pricing, border adjustments and the work of the Taskforce on Nature-related Financial Disclosures (TNFD) promise new asset classes in carbon, water and nature. And all that before we even get to the risks associated with depopulation trends seen in places like China, Japan and South Korea.
I don't want to spook the horses. But in exponential times, such events are likely to become commonplace. Our virtual youth advisory board is not there to solve them—only to grasp how change now cascades, something scientists are more used to seeing in ecosystems than in machines.
With all that on the table, here are our virtual panelists to help us kick off the discussion in the following three posts:
Each subsequent post will spotlight three of these six panellists, chosen for how they agree or disagree on a given theme. And it will also feature a seventh chair at the table, representing an even wider world.
Six young avatars, however well cast, share blind spots by construction. They are all human, all alive now, all speaking in the present tense about their own priorities. Yet the debates in this series turn on exactly the voices that this description excludes.
So, every session includes a seventh chair, whose occupant changes with the theme to represent whoever—or whatever—the six cannot. For the panel, it is a potential guardrail against becoming an echo chamber, providing the dissent that spotlights where some of our own blind spots hide.
As a dry run, let's put the trust question to three members of the panel. What criticisms do they have of the virtual advisory board approach?
Plenty, Kwame erupts. "You have built a version of me that will never ask you for money. That's very convenient for you. Every real conversation I want with people like you would be about capital and decision rights—and you have removed the only person who can walk out of the room when your answer is no."
Sofía objects from somewhere even more challenging. "Where I come from, people are killed for saying these things out loud. Journalists, land defenders—buried, not consulted. A composite voice like this can't be threatened, so it costs nothing to say what I say. Be careful about calling that a voice."
Then, Mei is something of a contrast: pragmatic and, on this question, perhaps the most disconcerting. "If the model produces better decisions, I do not much care where the input came from. You are asking about provenance. I am asking about output. Test it: run the panel, then run the real thing, and see whether the answers differ. If they do not, you have your answer. If they do, you might just have built a better model."
Marta, 39, youth-participation practitioner
For the first of the series' four runs, the seventh seat goes to someone with nearly twenty years of doing the real thing. Marta, forty-four, has convened youth advisory panels for two UN agencies, a development bank and four multinationals — and watched most fail on their own declared terms.
She is blunt about my virtual board. "You have built something that cannot embarrass you. I have sat in rooms where a nineteen-year-old took a chief executive apart, and the room went silent. That is not a data point. It is something happening to a person with power, in front of witnesses. Your board cannot do that — and that it is cheaper is exactly what worries me."
She is equally unsparing about the alternative. "Let me describe the real thing honestly, since nobody in my field likes to. We fly six young people to a European capital. They get a two-day workshop, a photograph, and a line in the annual report, which goes to a team that already signed off the strategy. The panel disbands when the funding cycle turns, often inside eighteen months."
What separates the real thing from the ritual is embarrassingly simple. "Pay them. Give them a seat with a vote, not an observer's chair. Multi-year terms, so they can watch what happened to their advice. And publish what you did about it — including what you rejected, and why. Do those four things and a real board is worth ten of yours. Do none of them, go virtual, and at least nobody flew!"
Her verdict is conditional. "The test is not whether your panel is real. It is what it displaces. Replace a good board and you have done harm. Replace nothing — because there was never going to be a board — and you have built a thinking tool. And if it tells you which real young people to bring next, and you use them well, it could be the best use of the technology I have seen."
It's genuinely too early to say. We will not resolve that issue here and should distrust anyone who claims we could. Over the next three posts, the virtual advisory board helps me scan the possibilities for upsides and downsides. We consider why time horizons matter, the geography and rules of the transition, and who—or what—is still likely to be disenfranchised by 2037.
These four posts are spinouts from my ongoing work on the Green Paper, 2037: The Sustainability Outlook, where I am lead author. This will launch in October 2026, with a White Paper to follow in January 2027. The opinions expressed here are mine alone.
Next in this series: Why 2030 Cramps Our Imaginations (due on 15-09-2026). And then Build Fast And Set The Rules (due on 22-09-2026) and Who—Or What—Still Can't Vote? (due on 29-09-2026).
This article is also published on John Elkington's Substack. illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.
illuminem briefings
AI · Ethical Governance
illuminem briefings
AI · Ethical Governance
illuminem briefings
AI · Corporate Governance
Associated Press
AI · Ethical Governance
Al Jazeera
AI · Ethical Governance
Deutsche Welle
AI · Nuclear