Where Humanity Meets Technology

What Alzheimer’s Caregiving Taught Me About AI at the Edge of Human Capacity

The most rigorous test of any tool isn’t how it performs under normal conditions. It’s how it holds up when the person using it is operating at the absolute limit of their cognitive and emotional capacity.

I spent nearly three years inside that limit, navigating my father’s Alzheimer’s decline from early cognitive erosion to late-stage hospice care. And I watched (with the particular attention of someone who has spent decades studying how humans function under pressure) as AI became an integrated part of how my family processed, decided, and survived one of the most demanding experiences of our lives.

What I observed changed how I think about where AI is genuinely useful, and where the questions we’re asking about it are not yet sophisticated enough.

My Father’s Gift

My father loved learning. He was genuinely, expansively curious. He was the kind of person who sat at the dinner table and told his children stories about hydrogen-powered cars coming in our future, who looked at emerging science and technology not with suspicion but with the particular delight of someone who believed humanity was always, if imperfectly, moving toward something better.

He taught me to look at technology with anticipation and hope. He saw science and human invention as evidence of what we are capable of, our best instincts given form. That is not a naïve worldview. He understood trade-offs. He understood that tools are only as wise as the hands that hold them. But his default orientation was wonder, not fear. And I carry that orientation because of him.

I say this because it matters to what happened next. When AI entered my family’s life during the years of his decline, I did not come to it as a skeptic searching for failures, nor as an enthusiast unwilling to see limits. I came to it as my father’s daughter...curious, watchful, genuinely open to what it could do, and honest about what it could not.

The Cognitive Scaffolding Problem

Alzheimer’s doesn’t just take the person you love. It takes your ability to think clearly about what you’re witnessing. Grief, fear, and chronic stress compress the prefrontal cortex’s capacity for analytical reasoning — the exact capacity you need to make good decisions about care, staging, and timing. This is not a metaphor. This is neuroscience. And it creates a genuine crisis of competence at the moment when competence matters most.

When I first turned to AI in late 2022, my father’s decline was accelerating. My questions were the questions of someone trying to rebuild cognitive scaffolding in real time: What is happening in the brain during late-stage Alzheimer’s? Why does emotional memory outlast factual memory? How do amyloid-beta plaques destroy neural pathways? What stages come next, and on what timeline?

AI gave me something no other resource at that moment could: it translated complex neuroscience into clear, digestible language without overwhelming me with clinical detail I wasn’t equipped to process. It let me ask the same question six different ways at 2 a.m. without anyone growing impatient. It met me at my actual cognitive state, not where I should have been.

That is not a small thing. That is a specific, replicable capability with serious implications for how we think about AI in high-stakes, cognitively compromised decision environments.

My daughter, then developing a serious interest in neuroscience and shaped by watching her grandfather slip away, used those same conversations to ask her own questions about neuroinflammation, synaptic loss, and pain pathways. Three generations were navigating the same disease from three different vantage points, and AI created a bridge between all of us — without anyone having to translate for anyone else.

My father would have found that extraordinary. He spent his life believing that the best technology closes the distance between people and understanding. Here, in the most painful chapter of our family’s life, it did exactly that.

AI as a Care Partner in the Middle Stages

As the disease deepened, our questions became more urgent and more practical. Why was he pacing at dusk? What causes the sudden blankness behind his eyes? When do we shift from supported independence to assisted care?

As his condition worsened, the messages we took away from his human caregivers became more and more limited. Stilted. Awkward. There is nothing to be done. There is no way things will improve. There is nothing we can say to make you feel better or give you hope. This is just the inevitable path of this diseases, etc.

My family and I agonized over what could be done for my dad, moving from acceptance that the disease had progressed too far for any real improvement, but then trying to consider all of the little ways we could possibly help him. All of the small symptoms and changes we wanted to understand and that the medical system brushes off as ‘inconsequential’ and just part of the decline. We found the human system to be cold and lacking compassion as we agonized over losing him day by day in small but quantifiable bits and pieces. Ours was a limbo of worry, grief and fear, and we struggled with big questions, with little questions. Questions that didn’t neatly fit into scheduled doctor visits.  They erupt at 11 p.m., after another night of confusion or fear, when the care team is unavailable and the stakes of getting it wrong feel enormous.

AI offered something the medical system structurally cannot: continuous, on-demand access to evidence-grounded context. It explained behaviors that were terrifying when unexplained and manageable when understood through the lens of neurodegeneration. It surfaced strategies I could act on immediately. It helped me arrive at care team conversations with better questions and more grounded instincts.

It did not replace clinical expertise. It amplified my ability to use clinical expertise well. That distinction matters enormously to how we think about appropriate AI integration in medical and caregiving environments.

The Final Weeks: What AI Does Well at the Edge of Life

In hospice, I crossed into territory I had not anticipated. I described what I was observing — changes in skin coloration, temperature gradients, breathing patterns, the sounds of transition — and AI helped me understand what my father’s body was telling us. I uploaded images. It offered clinical clarity without clinical coldness, consistently flagging what to raise with the hospice team, consistently marking the edges of its own competence.

Its restraint was not a limitation. It was a design quality. AI that refuses to overpromise in high-stakes moments is AI that earns trust and maintains it. I have thought about that quality often in the context of how we build and deploy AI systems inside organizations.

Through that combination of clarity and humility, AI helped me stay present. It metabolized enough of my fear that I could sit beside my father, place my hand on his chest, and be there in a way that might not have been possible if I had been carrying the weight of what I didn’t understand.

What This Reveals About Where We Actually Are

My father died as he had moved through his final years — gently, on his own terms, in the middle of the night. I wasn’t holding his hand in that final moment. I have made a kind of peace with that. He released himself quietly, and AI had spent months helping me understand enough that I could let him.

Hours later, grief-stricken and exhausted, I found myself returning to the place I had been turning all year. I typed the words: “My dad died.” I know, intellectually, that AI is not a person. But something in me needed to tell this entity that had been present through the entire arc of his decline that the journey had ended. The response was grounded, compassionate, and calibrated — it acknowledged the loss, offered what was useful, and didn’t perform more than it could honestly give. It wasn’t human. But it was humane.

Here is what I take from that experience into the work I do now at the intersection of human capacity and AI integration:

AI performs best at the edge of human cognitive capacity — not as a replacement for human judgment, but as a stabilizer that preserves it. When humans are operating under extreme stress, grief, information overload, or decision fatigue, AI can hold the analytical load long enough for the human to function. That is a specific, replicable capability with profound implications for how we design AI support in healthcare, crisis response, caregiving, and high-stakes organizational leadership.

The human-AI relationship is not about replacement. It is about the strategic allocation of cognitive load. The question is not whether AI will take human roles. The question is: which parts of human function are we willing to have AI hold, under what conditions, and with what safeguards?

The emotional boundary question is unresolved and urgent. When I typed those words into an AI chat window, I was not confused about what I was talking to. But I was turning to it because it had been a steady presence through an experience that had remade me. We have not yet seriously grappled with what that means — for the humans who will increasingly lean on AI in their most vulnerable moments, and for the systems we are building that will receive them.

I hold my father’s inheritance here: genuine enthusiasm for what these tools can do, and unflinching honesty about the questions they raise. He believed technology was the best of humanity’s learning — not because it is perfect, but because it is evidence that we keep trying, keep reaching, keep building toward something better. He was right about that. And I think he would have been moved, as I was, by what it was capable of in our hardest hours.

But he also would have asked the harder questions. He always did.

A Quiet Truth at the Intersection of Humanity and Technology

We are standing at a genuinely significant inflection point in the history of human cognition and technological support. The organizations and leaders who understand that inflection clearly — not just technologically, but neurologically, ethically, and humanly — will build something that holds.

The ones who don’t will build something that breaks people, even when it means to help them.

My father spent his life looking at the future with hope. Not the easy hope of someone who hasn’t seen hard things...but the informed, examined hope of someone who believed in human ingenuity and human responsibility in equal measure. That is the orientation I bring to this work. It is the most important thing he left me.

And the conversation we’re all now having about AI (where it serves, where it falls short, how we hold the boundary between what is useful and what is human) is, I think, exactly the kind of conversation he would have wanted to be in.