The Boom
Yes, I've had a number of occasions where I was literally astounded at what I'd been able to achieve working with generative AI. Never before have I had the ability (or imagined the possibility) of immediately rendering my own nuanced, deliberate intentions and achieving in minutes what days or weeks of collaboration with a number of deep specialists would accomplish—and only in theory!
Never had I experienced riding an experimental thought current to a cascade of conclusions, generating novel systems-level opportunities and insights that can transform the way we work and collaborate as organizations and teams.
Most of the field notes I currently have in mind are about novel workflows and frameworks, or the many possibilities arising from augmenting deep skill in research and human-centred design with generative and agentic AI processes.
My niche application for generative AI workflows is in the bookends of pre-analysis and post-analysis where I've identified significant gains for depth and structure, efficiency, and predictability. It would be dishonest to deny that this is the most lit up I've ever felt about experience design. Until I'm not.
The Doom
And yes, it's also true, that taking the plunge and learning, experimenting with AI was not a simple or light decision to make—and hasn't gotten any easier. Let's be real, as an experience designer of 12 years, who essentially commodifies empathy for business profit, I saw first-hand how harms are baked into this industry and system before generative AI ever emerged (a future field note in the works).
I was already struggling with whether I can ethically do my job in a system that disproportionately values expediency over ethics and evangelizes models of development like "move fast and break things" in the quest for ever expanding growth and profit. I was questioning what I was contributing to when the products or services I'm supporting only tolerate human-centred design when it's profitable—utterly let down by a dominant culture that rebuffs the intrinsic value of deeply understanding and respecting the people whose money and attention we seek, capture, profit from.
Put plainly: the root of all the harms we're seeing and experiencing with AI is systemic—when intrinsic value is transferred from individuals to dollars, what protects people and society from technologies and companies extracting wealth at all our expense?
The Tension
Is it any wonder then, that when OpenAI ushered in a new AI era, ecological and social devastation already came baked into this technology's promise and our collective achievements?
It certainly didn't come as a surprise to me when I saw the writing on the wall that the entire life I built was genuinely placed at risk if I maintained my refusal to engage with this full-spectrum AI emergence—especially as a senior research consultant.
I needed to choose between a) risking layoff and indefinite unemployment; b) dropping out of my profession and passion and restarting in mid-life with a child and a mortgage; c) learning to navigate this tension and committing to grounding my use of this technology in principles.
I chose c). Not because I found resolution but because I took a sober look at my own station and power. Refusing AI outright and cutting off my viability as a practitioner wouldn't stop AI. It would just stop me, my family, and our security.
Three Principles
Internalize AI's limitations
I've seen a senior leader earnestly believe they can get Copilot to analyze raw qualitative data as their favourite theorist and use that output as an evidence spine. These opaque LLMs are fantastic at coding, computing, carrying out directives, accepting specifications—but conduct a grounded analysis of any kind, they cannot.
Just as constraints are a launchpad for creatives, internalizing LLM constraints is a fundamental step to unlocking some truly novel and purposeful applications of AI.
Think first, start with outcomes
Immediately relinquishing your own agency and expertise to a model is a tragic first step to any work (or any step, really). That same leader didn't design the prompt they fed Copilot, they vibed it, which means countless implied fill-in-the-gap assumptions were woven into an already misguided use of AI.
Being deliberate means I walk back from an end goal and think through the structure and specifications of what I need to know or get done before entering a word into a chat. Once I'm oriented, I start reaching for the right tools for the job—who sees a hammer and thinks, what can I build using this?
Humans ARE the loop
I don't know about you but I've certainly reached my saturation point with this "keep humans in the loop" talk. Are we already at the point where the default isn't human intelligence and now the case needs to be made for our own involvement? Without deliberate intentions and specifications, an LLM isn't able to do meaningful work.
My own work with AI is focused on the pre and post-analysis bookends because that frees me up to focus more deeply on substance. When I feed my considered analysis into an intentionally structured generative AI workflow, I'm deciding how AI thinks and its place in the loop. Scaling this up to an organizational level is how to ground an LLM in institutional knowledge and trust it to interpret meaning—rather than simply prompting it to think like your favourite theorist.