A response to Uday Dandavate’s “The Synthetic Mirage”
Uday Dandavate wrote something important this week, and I want to engage with it seriously — not to refute it, but to extend it.
His core argument is one I hold as a practitioner: immersive field research is not a luxury or a premium add-on. It is the mechanism by which insight becomes conviction. When a design team sits across a table from a real person — a nurse describing her morning workflow, a procurement manager explaining why he didn’t buy — they experience something that no slide deck, no AI-generated profile, and no statistical composite can reproduce. The context, the hesitation, the off-hand remark that changes everything. That is irreplaceable. Full stop.
I agree with Uday. And I want to add something he didn’t.
The problem isn’t the tool. It’s the team’s situation.
Uday is right that the most dangerous use of synthetic personas is as a substitute for fieldwork — as a shortcut to false confidence. I have seen this happen with traditional personas for years: what starts as provisional scaffolding hardens into institutional truth, resistant to contradiction. AI makes this failure mode faster and more convincing.
But there is a different situation that Uday’s article doesn’t fully address, and it’s the one I encounter most frequently in my consulting work.
I work primarily with B2B and enterprise product teams. These teams are not choosing between good fieldwork and synthetic shortcuts. They are navigating a structural access problem that makes their situation genuinely different from the consumer research world.
Let me give you a concrete example. A researcher I supported waited eight months to get access to a particular type of clinical nurse. Eight months. The friction came from two directions simultaneously: the clinical schedules of the nurses themselves, and the account managers in her company’s sales team who were — understandably, if frustratingly — protective of their customer relationships. No amount of methodological commitment closes that gap. The team still had to make decisions in the intervening eight months. They still had to align on hypotheses, brief developers, challenge assumptions in sprint reviews, and pressure-test their thinking.
What did they have to bounce ideas off? Their own assumptions. Each other’s assumptions. Which is precisely the failure mode Uday is warning against — designing for a mirror of your own prior thinking.
What twins can actually do — and what they cannot
This is where I want to be precise, because precision matters here.
I have been testing a tool called Twyn — twyns.net — which builds digital twins of user profiles grounded in real research data: your own interview transcripts, available market data, sector-specific signals. We ran a white paper on its validity (mckayconsulting.dk/do-twyns-match-real-people/) in which we shaped a twin from part of a research interview, then asked it questions from the portion of the interview we had withheld. We compared its responses to what the real person had actually said.
The results were meaningfully better than generic, demographically-constructed AI personas — the kind that Uday’s critique, and the Aalto University research he cites, rightly describes as amplifying hallucination and bias. The difference is consequential: a twin built on real field data is not a replacement for fieldwork. It is a distillation of fieldwork — a way of keeping the intelligence from your last research round alive and available for the team to work with between the next one.
But I will not overclaim. Here is what the approach cannot do:
It cannot surprise you the way a real person can. The off-script moment, the hesitation, the contradiction that opens a new question — these emerge from genuine encounter and are not reproducible in simulation. A twin will not go somewhere you did not expect. A real person reliably will.
It cannot carry the context that lives outside the transcript. What you observe in a field visit — the environment, the body language, the moment when someone’s voice drops — is not captured in your interview notes. It lives in the researcher who was there. The twin can only work with what was recorded.
It cannot create the ownership that comes from shared experience. When a whole team does fieldwork together, the insight becomes shared property. They cannot unknow what they witnessed. A twin can inform a team’s thinking, but it cannot give them that.
These are not minor caveats. They are structural limitations that should shape how teams use this capability.
The right frame: between fieldwork, not instead of
The value I see — particularly for B2B and enterprise teams — is not replacing the field visit. It is making the time between field visits more productive.
Teams need something to think with. They need to be able to ask: “Would our clinical nurse user care about this workflow change, or are we solving our own problem?” They need to surface assumptions, challenge hypotheses, and avoid building on consensus-as-truth in the absence of fresh field contact. A well-grounded twin can play that role — not as an oracle, but as a structured provocation, a way of making the team’s assumptions visible and testable.
The diagnostic question I keep returning to is this: who carries the risk if the simulation is wrong?
If a team uses a twin to generate hypothesis-questions they will then take into their next field visit — the risk is low and the value is high. The twin is scaffolding for better questions.
If a team uses a twin to avoid the field visit altogether — the risk is severe and the value is illusory. The twin has become a mirror of existing assumptions, dressed in data-driven clothing.
The tool is the same. The practice surrounding it is everything.
A note on the B2B specificity
Uday’s world and my world are adjacent but different in ways that matter here.
Consumer research — the families in Mumbai, the patients in Detroit — is structurally more accessible. Researchers can recruit participants. They can design observation contexts. The ethical and logistical challenges are real, but they are not the same as trying to access a VP of Procurement at a mid-sized German manufacturer, or a specific subspecialty of clinical nurse who is employed by your customer’s customer.
For B2B enterprise teams, access is not a budget question. It is often an organizational and political one. Sales teams guard their relationships. Clinical approval processes move at their own tempo. Competitors protect their customers. The team has genuine need for something to think with between the moments when the field opens to them.
This is where I see the clearest legitimate case for what Twyn is trying to do — not as a consumer shortcut, but as a B2B team infrastructure for keeping user thinking alive between research cycles.
Where I land
Uday asks the design industry a hard question at the end of his article: are we building tools that make it easier to design for people, or tools that make it possible to design without them?
It is the right question. And my answer is that the distinction is a practice question, not a technology question.
The technology can be used either way. The responsibility for which way sits with the practitioner and the team — in the briefing they write, in the questions they choose to ask, in whether they treat the twin’s response as a final answer or as a prompt for the next field visit.
What I will say with confidence is this: a team that does real fieldwork, grounds their twins in that real data, and then uses those twins to stay sharp and aligned between research cycles is practicing more rigorously than a team that does fieldwork once, builds a persona deck, and then leaves it to calcify in a Confluence page for three years.
The field is still sacred. I am not arguing otherwise.
I am arguing that what B2B enterprise teams need — and what the best version of this technology can offer — is a way to honor the field between visits, rather than forget it.
Further reading and resources
📄 Uday Dandavate’s original article — “The Synthetic Mirage: Why Synthetic People May Be the Most Dangerous Name in Design Research” https://uday-dandavate.medium.com/the-synthetic-mirage-why-synthetic-people-may-be-the-most-dangerous-name-in-design-research-8897d1fc40d8
🔬 White paper — Do Twyns Match Real People? Our validity test comparing data-grounded twins to generic AI personas https://www.mckayconsulting.dk/do-twyns-match-real-people/
🛠️ Twyns — Digital twins for B2B product and service teams https://www.twyns.net
Michael McKay is a transformation and product-culture consultant at McKay Consulting (mckayconsulting.dk), based in Copenhagen. He teaches at the Royal Academy of Design and works with enterprise teams on user-centred capability building.
Leave a Reply