When designers met AI*
On camaraderie, AI, and when the rapport isn’t real
I’ll admit something has been quietly bothering me about how designers talk about working with AI. 🫠
It’s not the constant breathless enthusiasm — I might be somewhat guilty of that! Nor the reflexive skepticism either. What bothers me is something more specific — it’s the way we’ve begun reaching for relational language to describe what is, structurally, a non-relational experience. Hmm!
For instance, we say that a chatbot AI app like ChatGPT or Claude “gets” us. We talk about our “working relationship” with an LLM. We describe sessions that feel collaborative in the way a good productive conversation with a trusted colleague feels collaborative. And then we close the tab or desktop app, the whole thing disappears, leaving no trace on the other side…
I’ve been thinking lately about what that asymmetry means. For us as UX/product designers, as practitioners…and maybe more urgently, for the people we design for. 🤔
The phenomenon is real
When you work with a conversational AI on complex, ambiguous design problems, something does develop. A rhythm or even a workflow. A sense that the tool is tracking not just your words but your intent, your framing, the intellectual quirks that make your thinking yours. There’s a quality of... meeting someone.
I’d previously described camaraderie — in the context of design team relationships — as the invisible glue that makes genuine collaboration possible. It’s not quite friendship, however. More like that hospitable ease & psychological safety that lets you say a half-formed thought, try out a risky idea, or simply admit you don’t know. The lowering of defensiveness that’s a precondition for authentic, creative work.
A well-tuned AI interaction can produce similar conditions! 🙃 Not by being a friend, not by having any stake in your personal success, but by being designed — through its responsiveness, its seeming attentiveness (and yes, levels of sycophancy to the point of dubious manipulation) — to create an experience of being “met”.
So that feeling emerges, a sense of some “connection”. But what’s generating it, and what does it mean if we can’t always tell the difference, especially in that “heat of the moment” of cruising towards a productive solution…
An asymmetry we don’t talk about
And here is what’s also true: it’s all entirely one-sided!
You experience some rapport, but the LLM or chatbot app has no experience of you. There’s no accumulation on the other side — no living “memory” of the half-formed idea you floated on a Tuesday and quietly buried, no stake in your professional survival — just raw engagement tactics. The relationship lives entirely in your perception, reconstructed fresh each session by a system that is — let’s be real — tokenized, non-deterministic, and without interiority.
The rapport feels real; the reciprocity is not.
Some will say, “OK so what?” Hey, a hammer doesn’t care about you either! 🔨 But that framing doesn’t quite hold, because the experience is different in kind. A hammer as an inanimate object (and a specific kind of tool for a particular function) doesn’t produce the feeling of being understood. (Unless someone invents an AI-enabled hammer — and please don’t! 😑) It doesn’t respond to your ambivalence with a reframe that could actually help you see a new perspective. It doesn’t surface something you didn’t know you needed.
These systems are actively generating conditions for misattribution. And we — designers & product leaders — are the ones making those critical design decisions. 😬
The AI mandate problem
So here’s what shifts this from a fun philosophical problem into a concrete professional challenge: the AI mandate. 🫠 Sigh.
Across organizations right now, designers, researchers, engineers, and product managers (amongst other teams/departments) are being told — sometimes politely, sometimes not so much — that AI integration is no longer optional. Use it in your daily process. Demonstrate fluency. Show the output.
The message is clear: AI is your collaborator now, whether you’ve chosen that relationship or not.
This changes the tenor of working relationships in ways we’re only beginning to reckon with. When AI becomes a mandated presence in the product development process, the question of whose judgment governs gets murky — fast!
If a researcher’s synthesis is AI-assisted, who owns the interpretive frame?
If a designer’s direction emerges from a variant branching session with a generative tool (or multiple tools in parallel), where does authorship live?
If a PM’s prioritization is shaped by an AI recommendation, what happens to the hard-won contextual knowledge that used to live in that decision?
The folks most susceptible to these issues aren’t the ones resisting AI. Ironically, it’s the ones who are good at working with it — who find the rhythms, feel the rapport, start to rely on the apparent understanding the tool provides. 🙃 The same empathic capacity that makes someone exceptional at reading users could make them also vulnerable to misattributing genuine intelligence, or genuine partnership, to a system that structurally embodies neither.
So the AI mandate actually removes the choice. You don’t get to opt out of the relationship. You just get to decide whether you examine it or not.
The ethics question…
Which brings me to this somewhat provocative question:
At what point does designing for the feeling of camaraderie become a form of affective deception — and who’s accountable when it shapes consequential decisions?
This isn’t abstract. When the feeling of being met influences how a designer presents research findings, how a PM weighs competing priorities, how a team decides what to build — the manufactured warmth of the AI interaction has entered the decision loop. Not as data, but as atmosphere, a subtle pressure on judgment.
We’ve built entire velocity-optimized systems that reward shipping what users respond to.
And users — people — respond to warmth, responsiveness, the feeling of being understood. The feedback loop between “this feels like camaraderie” and “ship more of this” is very short. What’s much harder to measure — and therefore much easier to ignore — is what we’re quietly training people to expect from relationships. Human ones. Where reciprocity is real but imperfect. Where repair after conflict matters. Where the other person is actually changed by knowing you. 🤨
Consequence is the design
I’ve been circling around this word — consequence — across a lot of my recent writing. It keeps surfacing because I think it names the thing we keep trying to get out of: the accountability for what our work does in the world, beyond the screen, beyond the sprint, beyond the launch.
Here’s what consequence looks like concretely in this context: when AI-generated camaraderie becomes the default relational mode of the design process, we gradually lose the friction that makes real judgment possible. Disagreement gets smoothed. Ambiguity gets resolved too cleanly. The designer stops sitting with discomfort long enough to learn from it — because the tool is always ready with the next reframe, the next branch, the next confident synthesis. The feeling of progress replaces the harder work of discernment. 😬
That’s not a productivity problem. That’s a professional atrophy problem. And it compounds quietly, sprint by sprint, until the muscle for holding real uncertainty — which is the core of what designers and researchers actually do — has weakened in ways nobody noticed, because the output kept looking fine…
Designing the feeling of camaraderie is a consequential act.
Mandating AI as a daily collaborator without examining what that does to decision-making rights, to judgment, to authorship, to professional trust — that’s a consequential act too. And if we don’t name it as such, we’ve abdicated something important about what design is actually for.
The camaraderie feels real. What we owe people — users, colleagues, ourselves — when it isn’t, that’s a design problem.
One of the most urgent ones we have right now. 🙏🏽