Care used to begin in a waiting room. The fluorescent-lit, cough- and sniffle-filled, fill-out-the-clipboard-so-there’s-something-to-do kind of waiting room. Now, it often starts with a chat prompt on your phone: “How are you feeling today?” That sounds ridiculously simple. Except that it’s not simple.
Care that begins with a conversation instead of a commute feels different. I’m not talking about when a nurse has taken your vitals or a doctor has entered the examining room. I’m talking about before that. When you’re trying to figure out if you need to take the time to make an appointment. When you’re trying to decide if you’re being paranoid or prudent.
Sometimes, that chat prompt is being responded to by a chatbot. It asks you questions, and responds to your answers. It asks follow-up questions or encourages you to seek additional care. It “listens,” in the way that a chatbot can be said to listen. Sometimes it advises you to monitor your symptoms. Sometimes it suggests you go to sleep. Sometimes it tells you that you should get yourself to the emergency room. This matters. This matters in ways that people may not appreciate.
Health concerns rarely arise at a convenient hour.
They arise at 11:47 p.m., when you’re sitting on the couch in the dark, and you’ve already typed a few symptoms into Google and are now convinced that you have six different diseases and that all of them are lethal. So it’s nice to have something to talk to at that moment that can tell you what you might actually have, and whether you’re likely to survive until the morning.
And it’s nice to have something that you can talk to without needing to call an actual person, and talk to their voicemail, and wait for them to call you back. At 11:47 p.m., when you’re alone and scared, it can seem easier to talk to a bot. Less embarrassing. Less dramatic. You can admit more readily what you’re worried about in a chat box than you can in a phone call.
That’s true, by the way. Fear does funny things to our language. It minimizes. It obscures. It makes us say that we’re “probably overreacting” when what we mean is “I’m scared and I don’t know if I should be.” A bot won’t judge. It won’t seem impatient. It won’t make you feel like an idiot for asking one more question.
I don’t think this phenomenon is about “ replacing” doctors and nurses with chatbots.
I never have. I think that’s a straw man argument, a way to make the rise of chatbots in health care seem more exciting and controversial than it really is. Chatbots aren’t going to replace doctors and nurses. They’re going to fill in the gaps. And that’s where things get interesting.
What happens when a chatbot catches a condition that might have gone undiagnosed for another day or two? What happens when a patient who typically drops off the radar after discharge gets a text from a bot asking how he’s feeling and actually responds truthfully? Sometimes, the answer is “nothing.” Sometimes the chatbot will tell the patient to take two aspirin and call in the morning. Sometimes it will advise him to rest and drink plenty of fluids and to check back in if his symptoms worsen.
But sometimes, the chatbot will alert someone to a problem sooner than would have otherwise happened. And in health care, timing matters. In health care, timing is everything. That’s why the rise of chatbots in health care matters. Not because they’re the future.
Not because they’re high-tech. Half the time, they’re still a bit low-tech, if truth be told. But because they allow care to reach patients in places and times and ways that weren’t really accessible before. Sometimes, the front lines of health care aren’t hospital hallways. Sometimes they’re chat windows. Sometimes those chat windows are staffed by humans. Increasingly, they’re staffed by bots.
So what about the first generation of healthcare chatbots?
Well, they weren’t much. They weren’t special. They weren’t insightful. They weren’t some sort of dramatic, cinematic, life-changing, “I saw dead people” kind of experience. They were essentially glorified switchboard operators on steroids. “What time are you open?” “How do I cancel an appointment?” “Where is your parking?” “Can I get a text reminder?”
That kind of thing. Necessary, sure. But not exactly earth-shattering. Reminders. Simple logistical support. Basic symptom checkers that basically consisted of algorithmic flowcharts with a chat interface slapped on them like a cheap digital veneer. It was basically just a fancy online version of the brochure rack in the waiting room. Fine. But hardly memorable.
Still, that’s not to say they didn’t do some good. They shortened the phone queue. They reduced a bit of administrative burden. They gave people the ability to quickly and easily ask a question or two without having to hold for ages listening to music that had been compulsory since 1998. And that’s no small thing. Not because anyone was in love with an appointment chatbot. Please. But because it changed behavior. Quietly. Slowly. Almost insidiously. People started sharing bits of health-related data with digital technology.
Maybe it was appointment scheduling. Maybe it was insurance information. Maybe it was basic symptoms. It wasn’t much. But it was something. And once people started getting just a little comfortable sharing that data digitally, something important happened.
The way things change, right? Not with a bang. With a whimper. With a series of everyday, mundane actions stacking one on top of the other until, at some point, the mundane action didn’t feel so new anymore.
The first chatbots showed people they didn’t need to interact with healthcare just through the front door. They could do it online. On demand. In micro-interactions. A question here. An answer there. Not sexy. Not game-changing. But a critical part of something much larger. And yes. Sometimes those first chatbots were infuriating. You’d get 3 menu options in and find yourself getting bounced around between rote answers wondering if the computer was just playing some kind of joke on you.
That happened. A lot. But even in their clumsy, limited, sometimes infuriating way, they inched the relationship between the patient and the healthcare system a little closer toward something more sustained. And that was what mattered. Not the cleverness. The permission.
And this is where the FAQ chatbots fall apart.
However, the same framework is not applicable to healthcare. Here, the previous approach starts to fail.
Care cannot be reduced to a simple transaction with a human element that has to be controlled. Patients don’t show up with perfectly formulated questions and neatly described symptoms. They talk. They pause. They downplay. They say things like, “I don’t know, I just don’t feel right,” which is not a clinical description at all, but it might be the most accurate description they can manage at the moment.
A keyword chatbot can respond to “fever”. It can respond to “headache”. It can even handle “what’s my copay” pretty well. Okay. Good. That’s great. But how does it respond to, “I’ve been really tired for the past few days and I don’t know if I’m just being paranoid or what”? How does it respond to, “Sometimes my chest hurts but maybe I’m just being a hypochondriac”? How does it respond to the patient who keeps minimizing what’s going on because she’s terrified of the response?
This is what healthcare is. This is what it looks like. Not keywords, but ambiguity.
And handling ambiguity is where simple automation starts to get really weak.
Because care is emotional, messy, contradictory, incomplete. Sometimes patients don’t say what they mean. Sometimes they don’t mean what they say. Sometimes they’ll ask a question but really mean something else. Sometimes they don’t need a direct answer so much as they need some help figuring out what’s really going on. That isn’t scripted. That’s relational. Or the start of it.
To my mind, this is the real divide between a chatbot that can respond and a chatbot that can support care in a meaningful way. A response is mechanical. Any kind of understanding, even limited, changes the character of the exchange entirely.
And character matters. More than some technologists like to acknowledge.
Because a patient who feels like she’s been heard is more likely to engage. More likely to follow up. More likely to be honest the second time if the first go-round was handled compassionately. That isn’t touchy-feely. That’s practical. Engagement drives better outcomes. It keeps patients from just drifting off and hoping the problem will somehow resolve itself. Which is a very human response and a terrible medical strategy.
So the healthcare folks started asking different questions. Not just: can it answer the question?
But: can it adapt? Can it pick up on patterns? Can it tell when someone is actually more concerned than they’re letting on? Can it follow a conversation over time? Can it escalate without freaking patients out? Can it reassure without blowing them off? Can it sustain a conversation long enough to be helpful not just efficient?
That’s where the limitations of the first generation lay. Pure automation could only go so far.
What was needed next was something smarter, yes, but also something more human in all the ways that actually mattered. More fluid. More contextual. Better at riding the twists and turns of an actual conversation without dragging the patient back to the decision tree every time they freewheel off script.
Because healthcare isn’t about answering questions. It’s about noticing when someone is trying to tell you that something is wrong even when they don’t know how to say it yet.
If you’d like, I can continue this piece in the same voice with the next sections, such as AI triage, chronic care monitoring, mental health chatbots, and the risks and limits of chatbot-led care.
The Importance of Emotional Design in Healthcare AI
Let’s not underestimate the role of data. It’s crucial. Numbers are crucial. When your symptoms started, what drugs you’ve been taking, whether your pulse rate is up, all of this is crucial. But healthcare isn’t just a series of numbers dressed in scrubs. It’s intimate. It’s patients turning up exhausted, terrified, humiliated, bewildered, or trying like hell to sound more composed than they feel.
Which is why the tone of a digital interface is so important.
Really important. Not just some feel-good branding exercise important. “Please tell us your symptoms” is not the same as “Hi, we’re here to help. What’s going on?” Functionally, the two are the same. But emotionally, they land very differently. One is akin to filling in your tax returns while unwell. The other is like someone has left a door open for you.
This distinction impacts behaviour.
Patients open up if they feel safe. They fill in more details. They don’t censor themselves quite so rigidly. They add in the thing they were just about to omit. In healthcare, the thing someone is just about to omit often ends up being the thing that proves most vital. A slight pain in the chest. Dizziness that we laughed off. A sentence that begins with, “This is going to sound silly, but…”
It doesn’t sound silly. It sounds human.
So when we talk about emotional design in healthcare AI, we’re not talking about some tinsel on the Christmas tree, like changing the font in a patient portal and calling it empathy. We’re talking about specific decisions around how a system communicates, when it pauses, how it handles distress, and whether it can tell the difference between a routine chat and someone quietly falling apart at 1 am.
That’s not trivial. That’s everything.
A well-designed healthcare AI system doesn’t just ask questions. It makes space. It registers if someone’s words carry a hint of panic. It doesn’t issue orders like an overconfident GPS. It backs off if the person at the other end sounds like they’re struggling to cope. It might ease up on the tempo. Use simpler words. Reassure before moving on. Provide one suggested next step rather than five. Frankly, sometimes that’s half the battle won right there.
Can AI be empathetic?
No, not really. I don’t think we should pretend otherwise. It doesn’t actually care about you. It won’t worry about you over a cup of coffee. It won’t sit with the emotional weight of what you’ve told it. But can it mimic the signs of listening well enough for people to feel safer continuing? Absolutely. And however much that might make some people uncomfortable, it doesn’t change the fact that it matters.
Because if people feel like they’ve been listened to, even by a machine, they’ll generally tell you more. That’s just how human beings operate. Not always, no, but enough for it to make a difference. A patient who feels like they’ve been brushed off will clam up. A patient who feels like they’ve been met with some degree of care, even simulated care, will more often stay in the conversation a little longer. Long enough to raise the issue that’s really worrying them. Long enough to answer more truthfully. Long enough, perhaps, to get the help they need a little sooner.
That’s the fork in the road here, in my view.
Healthcare systems that ignore emotional design will remain purely transactional. Quick, perhaps. Efficient, perhaps. But also cold. Forgettable. Easy to abandon. The systems that take emotional design seriously will foster trust, and trust in healthcare is not some vague optional extra. It’s a currency. It’s a door opener. It’s the difference between someone engaging with the process and someone slipping quietly out of the back door because the experience felt too clinical, too hasty, too uncaring.
And let’s be clear, nobody wants to funnel their deepest fears into an interactive spreadsheet.
Long-Term Memory and Patient Engagement
Now take that same concept and extend it over time. Imagine a healthcare system that recalls you about as well as a vending machine recalls the person who last pressed B7. It works. It delivers. Technically, mission accomplished. But emotionally? It falls flat. You log in, start over, repeat yourself, re-explain the medication change, mention the same side effects again, correct the same missing context, and by the end of it you’re not just tired, you’re slightly irritated. Fair enough too. Few things make a person feel less cared for than having to retell their own health story from scratch every single time.
Now imagine the alternative. The system recalls that your medication changed last month. It recalls that you had nausea for a week after. It knows your sleep has been patchy. It knows that Mondays are generally rougher for you, or that your symptoms tend to dip after treatment days. Suddenly the experience feels different.
Not magical. Not human, exactly. But connected. A line instead of a series of dots. “Has the new dosage helped you sleep any better?” comes across very differently from “Enter your symptoms.” One sounds like a follow-on. The other sounds like customer service with a pulse monitor attached. That continuity matters because people want to feel known.
Especially in healthcare, where so much of the experience already makes you feel reduced to forms, scans, values, percentages, and those jaunty little patient portals that somehow manage to make even serious concerns feel like you’re filling out a product return.
When a system recalls context, it reduces friction. But more than that, it reduces emotional friction. You don’t have to psyche yourself up to explain everything all over again. You can start where you actually are. And that makes engagement a little easier. Not dramatically easier. Not foolproof. Just a little bit easier.
Of course, this is the point at which privacy issues come marching in, and rightly so. They should. They absolutely should. Long-term memory in healthcare AI is not a feature to be thrown around lightly. It involves serious questions about consent, data limits, storage, access, misuse, and who can know what. If a system can remember more, then it’s responsibility has to increase with it. No shortcuts there. No quick “move fast and break things” shenanigans here, thank you very much. This is healthcare. Not a food-delivery app.
Still, from a patient engagement perspective, personalization is powerful. People return to systems that feel familiar. They respond more freely to instruments that reflect their own history back at them with some degree of care and accuracy
. And in a domain as personal as health, to be remembered can feel oddly emotional. Not because people confuse software with affection, obviously, but because simple recognition can be comforting. To be remembered is to not have to start from scratch. And when you’re already exhausted, in pain, anxious, or simply fed up, that’s not nothing.
In Conclusion: the Frontline Goes Digital
Care isn’t being pushed out. It’s being extended. That’s what I think people misunderstand, anyway. They see digital care as a kind of retreat from human care, as if care is just sort of backing away, smile and wave good-bye, now I’m going to go let a chatbot handle it. I don’t think that’s what’s going on at all. I think it’s more nuanced. And also just more interesting.
The doorway is moving. And care isn’t just starting at the front desk, or the clinic door, or even the waiting-room fluorescent light that’s just so demoralizing, you know? It’s starting on screens. It’s starting in messages. It’s starting with a little text exchange that’s very simple, and very human-sounding. A question. A check-in. A reminder that comes before it’s a problem.
That doesn’t replace the clinician. And shouldn’t. I’m the first one to say that if you try to sell that idea, I’m going to be the first one to complain. But it does expand the perimeter of care. It does expand support outside the visit. It creates some continuity between visits. It gives the provider some earlier warning signs, and more context, and more opportunity to prevent something from getting out of hand before it does. It gives the patient a way to have a voice before they wouldn’t have had a voice at all.
And sometimes, you know, that is what the difference is between catching it early and having to clean it up later.
No, I don’t think this is humans versus automation. That seems like a really reductive way to frame it. The question is, can automation expand our ability to care without dehumanizing it in the process? Can it take on some of the burden without making it feel impersonal? Can it make care systems feel more present, rather than less?
That’s the challenge. That’s the opportunity, too. Healthcare is, at its core, a deeply human endeavor. It will always be. The fear is human. The relief is human. The uncertainty, the hope, the little moments of courage that people have to find in order to reach out and ask for help in the first place, that is all human. AI isn’t going to replace that. It’s just going to learn enough of the language to try and be a little more helpful.
