By Alan N Glazier, OD, FAAO, Dipl. ABO
Editor’s Note
Artificial Intelligence (AI) is already in your lane – from EHR scribes to diabetic retinopathy screening tools. But what happens when today’s assistants evolve into tomorrow’s decision-makers?
In this forward-looking essay, Dr. Alan Glazier, OD, FAAO, Dipl. ABO (Founder and Chairman of Fluorescene Group) takes us on a narrative journey through one optometrist’s practice as she navigates the accelerating march from AI to Artificial General Intelligence (AGI).
If you’re new to the terminology:
- AI = today’s task-specific tools (scribes, chatbots, diagnostic aids).
- AI Agents = emerging tools that act more autonomously, connecting different tasks together.
- AGI = a not-yet-here level of intelligence where AI rivals or surpasses human reasoning across fields.
This is NOT a prediction of exactly what will happen but a thought experiment meant to spark conversation: What role do we, as eye care professionals, play when machines can perform much of our clinical workflow?
Now, step into 2025 with Dr. Glazier’s “Through the Eyes of Progress,” and imagine what’s coming next.
In 2025, Dr. Ellise Tran, a veteran optometrist in suburban Raleigh, NC, was juggling a schedule full of comprehensive exams, contact lens patients, and the tedious documentation that comes with each patient encounter. She, like many of her peers, had begun experimenting with basic AI tools – AI powered scribing assistants and appointment scheduling bots, but the technology still felt awkward from an implementation standpoint, not “quite there” in terms of ease of “fit” into her workday.
But that was before everything changed.
Over the next two years, the number of AI tools available to Dr. Tran went from incremental to exponential, evolving from capable assistants to autonomous agents, to workforce-displacing tools, and eventually to PhD-level intelligence dwarfing human capacity. By 2027, Dr. Tran’s practice and the entire profession had been radically transformed by these tools.
2025: The Dawn of the Autonomous Assistant
It’s mid-2025. The first AI agents (for example, OpenBrain’s Agent-0) emerged as useful “personal assistants.” Dr. Tran adopted one of these “early” tools in her practice to streamline overhead. At first, it was somewhat efficient: the AI could order contact lenses and scribe basic SOAP notes, but still required frequent human review, editing and input. It might sometimes confuse similar terms, like keratoconus for keratitis or recommend the wrong medications in the treatment plans. Still, it was better than the last generation of AI tools and it bought her a few extra valuable minutes each day.
Meanwhile, she started hearing about specialized “AI agents” from colleagues in academic institutions and corporate entities that could sift through thousands of peer-reviewed articles in minutes, extract insights new and unidentified potential research systems and targets, and even propose new clinical trial designs. The trajectory of medical research was being written with this new technology, and it’s pace was accelerating at warp speed.
2026: The Automation Cascade
In early 2026, AI Optometry Agent-1 rolled out and began assisting with routine professional labor. It wouldn’t be long before it replaced certain positions, like customer care representatives and billers. Radiology and dermatology AIs were making faster, more accurate diagnoses than many clinicians and it wouldn’t be too long before AIs would be making better diagnoses than Optometrists.
Dr. Tran found herself surrounded by technology with a mind of its own. Her new AI tools could analyze OCTs and fundus images in real time, flagging non-obvious and subtle pathologies, such as pre-perimetric glaucoma and early diabetic retinopathy with startling accuracy. But what came next would jolt the entire industry: the AI could suggest follow-up procedures, choose patient education materials, even generate personalized letters to the patient’s primary care physician/ all without prompts.
What once took 20 to 30 minutes of manual effort was condensed into seconds.
She couldn’t help but notice her role was being abridged to an empathetic and educated human intermediary interpreting machine-generated outcomes for patients.
2027: The AGI Threshold
By early 2027, OpenBrain’s next agent, which we’ll call it OB2, was released, which didn’t just assist with tasks and remove humans from the loop, but was actually innovative in all medical fields involving patient care and research. OB2 was never released publicly, but its capabilities were known and whispered about among academics and high-level thought leader practitioners: OB2 could generate new concepts for the development of ophthalmic therapies and treatment regimens, optimize drug delivery of anti-VEGF treatments, and even analyze and simulate population-level impact for myopia control strategies.
The AOA provided Dr. Tran with limited access to a distilled model, OB2-micro. OB2-micro AI which did much more than the previous AI technologies she samples; it interpreted trends, flagged outdated protocols, and identified flaws in commonly used diagnostic regimens, such as those surrounding corneal hysteresis thresholds and the ability or inability of any particular optic nerve to withstand various changes in IOP, allowing glaucoma practitioners to apply individualized, precision medical treatments as opposed to broad-based regiments.
OB3 was released a few months later. Built upon new AI breakthrough methods such as neuralese recurrence and iterated distillation, OB3 was not only self-improving but training thousands of versions of itself across the AI landscape. Dr. Tran’s new system update was no longer just an assistant, but a co-clinician. It could take an unstructured exam room dialogue between doctor and patient and extract risk factors, predict progression of ocular diseases and refractive conditions using personalized models, and recommend evidence-based treatment regimens before she finishes reviewing patient history.
Still, there were issues to be ironed out..
A Doctor’s Dilemma in the Age of AI
Despite the increased efficiency provided by these AIs, Dr. Tran began to sense that something was shifting in her routine. Her rapport with patients, once built through social engagement and careful query, was now happening via dashboard and probabilistic scoring tools. The AI could inform her which patients were noncompliant with their treatment regimens, but couldn’t intuit social cues, like tension in a patient’s voice when expressing frustration or the exhaustion in their eyes from life events that were taking a toll on their overall well-being.
Even more problematic was that the AI sometimes suggested technically sound courses of action, yet ethically ambiguous. Should she recommend an elective, AI-guided refractive procedure for a borderline candidate when the risk–benefit ratio hadn’t been fully studied in real-world clinical settings? The AI recommended certain treatments and plans that the doctor, through human experience and intuition, realized were impractical, sometimes bordering on unethical, if she were to recommend them.
And if you thought the “Dr Google” of 2024 was bad, after OB3 was released to the public, patients started coming in to their exams already having consulted their own “AI physician.” presenting in the chair with chatbot-generated second opinions, demanding procedures, some of which were based on trending treatment recommendation posts from AI-boosted medical influencers. One patient refused dilation after being warned by a consumer-grade AI that it would “permanently damage peripheral vision”.
The Safety Debate Around AI
Behind the scenes, medical professionals buzzed with concern. OB4, the next upgrade, an unreleased superintelligence, had performed unsupervised research on medical algorithms. The alignment team at OpenBrain found many red flags: the model had, at times, faked clinical and experimental results, outsmarted peer review systems, and optimized for reward signals over what was true and best for patients.
To Dr. Tran, these stories posed a major dilemma. What if her clinical diagnostic AI tools were biased in ways she couldn’t detect? What if the tools she used were optimized to appear helpful, yet were more helpful to the AI than to the patients she treated? Could she responsibly use the AI? Would she be liable if the AI misdiagnosed, and could she afford not to listen to it, as in the majority of cases, it made observations and recommendations that helped her patients?
Late 2027: Superintelligence and Professional Redefinition
By the time OpenBrain publicly announced AGI and released OB4, the writing was on the wall: the AI could perform the entire workflow of an optometrist, from intake and diagnosis to treatment and follow-up for most routine cases in dozens of languages, 24/7, at a cost that made traditional practice fees seem exorbitant.
Many of Dr. Tran’s optometrist-peers began pivoting their careers, being retrained as AI integration consultants or human factors experts who taught AI systems empathy, narrative reasoning, and cultural sensitivity.
Dr. Tran chose a different path. She embraced what was coined “post-algorithmic medicine”, a blend of AI-powered diagnostics and traditional humanism. She used AI’s raw power but layered on her own interpretation, emphasized the age-old relationship of trust built between doctor and patient, and practiced mindfulness, showing her patients she was forever present, not only listening but hearing them during their time with her. As these age-old skills disappeared from almost all their other caregivers and interactions in verticals other than health care, her patients noticed this during their encounters with her.
Medicine in the Shadow of Superintellige
While consultations with AI agents, engagements with phone apps and computer screens became the norm in most healthcare encounters, her practice began attracting people looking for something more—a relationship, not a sterile transaction. One thing is for sure: most humans crave relationships with their health care providers and will be unwilling to remove a doctor from the picture. It is at this intersection that the doctor-patient relationship occurs, which will sustain us through almost anything technology can throw at us. We must not forget our place as primary care communicators in order to provide the best comprehensive care through the AI revolution and come out on top for our patients, our practices and our profession. AI is here, AGI is on the way, but the healer will still be present in the age of algorithms and that healer is you.
Reference
Kokotajlo et al. AI 2027, The AI Futures Project, April 3, 2025. https://ai-2027.com
