What Diabetes Fears Most Is ‘Feeling Fine’: How AI Became My Around-the-Clock Health Sentinel

By Michael C.S. So | AiX Society

By then I had already been diagnosed with type 2 diabetes. Fifteen years ago, at the age of thirty-three, I was the typical Hong Kong “diabetic office worker”: no family history, simply years of staying up late, irregular meals and heavy drinking at business dinners. By height and weight my BMI was over thirty, classed as severe obesity. After the diagnosis I took medication for a while, then, feeling there was nothing seriously wrong, stopped on my own. That step cost me far more than I imagined at the time. Over the years I was off medication, the side effects of high blood sugar quietly accumulated: irritability, lethargy and blurred vision whenever my glucose spiked. It was only when that wound on my back appeared and a week of antibiotics made no difference that a blood test revealed my glucose had shot past twenty mmol/L. The doctor put it bluntly: the situation was critical, and it had not been caused in a single day.

Today, under medical guidance, I am back on treatment. My regimen now includes an SGLT2 inhibitor, and my daily pill count has fallen from more than ten back then to four, with my weight also back where it should be. I was given the second chance that many patients never get.

I share this because it has been my entry point for thinking about AI over these past few years. As an AI consultant, I long treated AI as no more than a tool that “writes and analyses”. It was only when I had to manage this condition over the long term that I began asking a serious question: now that AI can see, hear and remember what we say, can it do more than finish my work for me — can it help me protect my health? My answer is yes, and it is affordable for every person with diabetes. Below are four uses I have either tried myself or consider genuinely practical.

First, use AI’s vision to “keep an eye” on your own body. If I had photographed that wound on my back back then and let AI compare its healing progress every few days, it would have raised the alarm long before the antibiotics failed and I finally took a blood test. Vision models are ideally suited to the kind of tracking that “the naked eye is too lazy to do daily”: wound size, the extent of redness and swelling, and changes to the skin, all turned into an objective record. The same logic applies to diet — photograph each meal and ask AI to estimate the carbohydrate content and glycaemic load, which is far more accurate than relying on “I think I ate less”. And then there are the test reports: most patients cannot read abbreviations such as eGFR and UACR. Photograph your report and ask AI to explain in plain language “what stage my kidney function is at and how far I am from the danger line”, which is more proactive than waiting until your next follow-up to ask the doctor.

Second, use voice AI to log your mood and routine, and let it spot the patterns for you. The irritability and fatigue I felt during those years off medication were in fact signals of blood-sugar fluctuation, but I was completely unaware of them. Now, before bed each night, I tell a voice assistant a few lines: “I only got to sleep at two this morning, had three meetings, felt restless and ate two bowls of rice at dinner.” AI can stitch these fragments into a trend and remind me that “your late nights and mood swings have been high this week, so your glucose may be unstable”. It does not need to be a doctor; it is simply a note-taker that never sleeps and is willing to listen to you grumble, reflecting back the self you cannot see.

Third, use AI to translate medical information into action. The professional threshold of diabetes management is high: terms such as SGLT2 inhibitors, RAAS inhibitors and cardiorenal protection are hard for ordinary patients to digest. I feed what the doctor says and the instructions on my medication packet into AI and ask it to explain in three sentences “what benefit this drug gives me and when I should see a doctor” — the equivalent of carrying a patient health-education nurse with me at all times. Research shows that early use of drugs with cardiorenal protective effects can delay progression to end-stage kidney failure by up to fifteen years — but only if you understand and remember what the doctor has told you.

Fourth, use AI to safeguard medication adherence and resist the urge to “stop because I feel fine”. The main reason I stopped medication back then was that the old regimen meant more than ten pills a day, which felt excessive and troublesome. Today the count is down to four and the burden is lighter, but the real enemy remains that thought of “I’m fine now”. What AI can do is gently remind you at a fixed time each day, turn “taken today” into something that needs no thought, and, when you miss doses repeatedly, remind you of the consequences using your own past data — rather than a cold, indifferent alarm clock.

Here I must make one thing clear: AI is an assistant, not a doctor. It cannot diagnose you, cannot decide your medication for you, and should never make you feel that “with AI watching, I don’t need to see a doctor”. Quite the opposite — its greatest value is that, by the time you see a doctor, you already arrive with your own observations and data in hand.

This matters especially in Hong Kong. The figures from the Hong Kong Diabetes Association are sobering: one in five adults with diabetes develops the condition before the age of forty, with an average age of onset of just thirty; and one in three people with type 2 diabetes will develop diabetic kidney disease. It is estimated that over 200,000 people in Hong Kong face this hidden threat. The most dangerous part is that in the early stages there is often “no discomfort” at all. The Association therefore urges people with diabetes to have a “one blood, one urine” check at least once a year — a blood test for eGFR and a urine test for UACR to monitor kidney function. I have written both into my phone’s annual reminders, and from this year I will first photograph my report and let AI read it through for me before I take it to the doctor.

This year marks the thirtieth anniversary of the Hong Kong Diabetes Association, and they have produced a health-education short film, Kidney Beans, about precisely this misconception that “no discomfort means nothing is wrong”. I sincerely recommend you spare ten minutes to watch it: diabetes-hk.org/kidney-beans/

Looking back, if I had had today’s AI tools fifteen years ago, I might not have wasted those years off medication. Technology cannot save every procrastinator, but it can help those who take themselves seriously to spot danger earlier. AI is not just for writing articles — used well, it can be the teammate that is always online and never tires of your worries in the fight against chronic disease.

Originally published in Chinese on HK01 (香港01). Read the original: https://www.hk01.com/article/60389051.