Beyond the Algorithm - AI, Chronic Care, and the Case for a Computer-in-the-Group Approach

Executive Summary

Chronic disease has outgrown the systems built to manage it. Diabetes, hypertension, cardiovascular disease, and chronic kidney disease now drive a large and rising share of mortality, disability, and long-term cost across both high-income and low- and middle-income settings. Yet most health systems remain organized around episodic encounters, fragmented specialties, and institution-centric workflows that are poorly suited to lifelong care.

The consequences fall hardest on rural, peri-urban, and underserved communities, where a recurring three-part failure plays out: people do not know they are at risk, do not know what to do after diagnosis, or cannot access affordable, coordinated care. This is not only a clinical failure. It is informational, behavioral, geographic, and relational.

Artificial intelligence can help. Evidence increasingly supports AI for risk stratification, remote monitoring, conversational support, symptom triage, and chronic disease self-management. But AI cannot, by itself, fix structural deficits in workforce, financing, trust, access, and accountability. The right question is not whether AI can replace clinicians. It is how AI can be embedded inside human systems so that people, providers, programs, and payers work together more intelligently.

That requires moving beyond human in the loop to computer in the group. Human-in-the-loop is a critical safety pattern. But chronic care is delivered by groups — patients, families, nurses, community health workers, physicians, employers, insurers, clinics, hospitals, and public programs — that are often disconnected, inconsistent, and overwhelmed. Drawing on Thomas Malone’s work on collective intelligence at MIT, this paper argues that AI should be designed as a participant in coordinated human systems, not as a parallel authority that humans supervise after the fact.

A horizontal engagement layer can close that gap. This is the role SabaiHealth plays: a persistent, AI-powered care companion that sits across institutions — delivered through the messaging apps people already use (LINE, WhatsApp, Telegram) — helping the chronic care ecosystem identify risk earlier, maintain continuity between visits, personalize support, and route people to the right human and institutional resources at the right time.

The Issue at Hand: Chronic Disease Has Outgrown Legacy Care Models

The global burden of chronic disease has reached a point where it is no longer merely a medical issue; it is an organizing challenge for healthcare systems, economies, and societies. Diabetes, hypertension, cardiovascular disease, chronic respiratory disease, and chronic kidney disease are among the leading causes of death and disability worldwide, and the epidemiological centre of gravity has shifted sharply toward Asia and the developing world.

The challenge is intensified by multimorbidity. Diabetes and hypertension frequently coexist and act as gateway conditions into chronic kidney disease and cardiovascular disease, making them especially important entry points for prevention and long-term management. Global analyses show that chronic kidney disease attributable to type 2 diabetes has more than doubled since 1990, while CKD related to hypertension has risen sharply, reflecting the compounding effects of poor detection, limited control, and fragmented care.

The Scale of the Challenge

• India: NCDs account for roughly 60% or more of all deaths, with hypertension, diabetes, CVD, and chronic respiratory disease contributing substantially.
• Thailand: NCDs account for approximately 7 in 10 deaths. CVD, diabetes, and chronic respiratory disease remain major national priorities. Kidney disease is closely tied to the country’s high burden of diabetes and hypertension.
• Malaysia: CVD is the leading cause of death. Diabetes, hypertension, and high cholesterol remain highly prevalent, with large numbers of adults living with multiple NCDs simultaneously.
• Global: CKD attributable to type 2 diabetes has more than doubled since 1990.

The systems response has lagged behind the epidemiology. Most healthcare delivery models were designed for acute episodes: diagnose, intervene, discharge, follow up. Chronic disease does not behave that way. It unfolds over years, requires sustained behaviour change, depends on continuity and adherence, and often worsens in the spaces between appointments rather than inside them. A person with diabetes and hypertension does not need help only during a clinic visit. That person needs guidance when deciding what to eat, remembering medications, interpreting symptoms, understanding lab trends, navigating side effects, or deciding whether swelling, fatigue, or chest discomfort is urgent.

This gap between how chronic illness behaves and how care is organized creates a recurring three-part failure pattern:

1. Invisible risk: Many people do not know they are at risk, because conditions such as hypertension, early diabetes, and early CKD can remain asymptomatic for years.
2. Lost in navigation: Even after diagnosis or an abnormal screening result, many people do not know whom to ask, what the finding means, or what to do next.
3. Inaccessible care: Even when people understand the need for care, they may lack timely access to affordable, coordinated services — especially if they live far from specialists, depend on overstretched primary care, or must navigate fragmented public-private systems.

These failures are not distributed evenly. Rural and underserved populations face a double burden: equal or higher chronic disease risk, combined with lower access to screening, follow-up, health education, and specialist care. Distance, transport costs, time away from work, digital exclusion, low health literacy, and limited continuity with a single provider can all turn manageable chronic conditions into expensive and devastating late-stage disease.

Where AI Comes In

Artificial intelligence is relevant to chronic care because chronic care is, at its core, an information and coordination problem as much as a clinical one. It requires repeated detection of subtle changes, timely interpretation of distributed signals, and personalisation of support over time. These are areas where machine learning, conversational systems, and intelligent automation can make a meaningful contribution.

Early risk detection and stratification

AI models can identify patterns in routine health data — claims, lab results, questionnaires, wearable streams — that suggest elevated risk for diabetes complications, uncontrolled hypertension, CKD progression, or cardiovascular events. In settings where data are sparse, even lightweight systems combining simple intake questions, known risk factors, and conservative rules can improve awareness and prompt earlier screening.

Continuity between visits

Chronic disease outcomes depend less on what happens in a 10-minute consultation than on what happens in the following weeks and months. AI-enabled conversational systems can reinforce medication adherence, support health education, encourage habit formation, collect symptoms, and respond to common questions in ways that extend support beyond the walls of a clinic. Remote monitoring supported by AI can also identify deterioration earlier and trigger follow-up before a crisis emerges.

Prioritisation

Every health system has finite attention. Hospitals, clinics, community health workers, and employers all need to know who is stable, who is drifting, and who needs immediate escalation. AI can help sort large populations into meaningful risk groups, allowing teams to focus on the people most likely to benefit from outreach or intervention. In this sense, AI is not just about prediction. It is about helping scarce human capacity land in the right place.

Navigation

Many people with chronic disease do not need a diagnosis from AI; they need clarity. They need to know whether a symptom may be urgent, whether they should contact a clinic, whether they should renew medication, whether a side effect is expected, and which part of the health system is appropriate for their next step. Properly designed AI can support these “what now?” moments, especially when tied to local care pathways, conservative escalation logic, and clear handoffs to human teams.

In short, AI is most valuable in chronic care when it does not try to act as an isolated substitute for medicine, but rather as a scalable layer of vigilance, interpretation, and engagement.

Where AI Cannot Come In Alone

The enthusiasm around AI can obscure a basic truth: chronic care is not just a data problem. It is a social, institutional, and moral problem. People do not fail to control diabetes or blood pressure only because they lack predictions. They also fail because food environments are unhealthy, transportation is unreliable, medicines run out, care plans are confusing, trust is fragile, and life circumstances constrain what is possible.

AI cannot solve those structural deficits by itself. It cannot create nurses, nephrologists, or community health workers where none exist. It cannot guarantee affordable medication, build roads to rural clinics, or repair fragmented reimbursement arrangements between public and private sectors. It also cannot assume away inequity in the data used to build it — a large share of healthcare AI development still relies on high-income, tertiary-care, or urban datasets, which raises the risk of underperformance in exactly the populations most affected by access gaps.

There are also limits of judgment. Chronic care decisions often involve trade-offs between clinical benefit, side effects, affordability, family context, health beliefs, work constraints, and patient preferences. Multimorbidity further complicates the picture, because what is ideal for one condition may be undesirable or harmful in the presence of another. These are not merely computational decisions. They require conversation, trust, accountability, and shared decision-making.

A further limitation is legitimacy. Patients do not ultimately place trust in an algorithm; they place trust in a care relationship and in institutions that stand behind decisions. That is particularly important when the stakes are high, symptoms are alarming, or recommendations are difficult to understand or accept. AI can support confidence, but it cannot replace the need for an identifiable human and institutional chain of responsibility.

The implication is not that AI is marginal. It is that AI must be placed properly. In chronic care, the right role for AI is not sovereign decision-maker. It is disciplined collaborator.

From Human-in-the-Loop to Computer-in-the-Group

Human-in-the-loop has become the default language for safe healthcare AI, and for good reason. Across diagnosis, triage, monitoring, and support applications, research indicates that systems combining AI with clinician or trained human oversight improve accuracy, reduce errors, and increase trust compared with either fully manual or fully automated approaches. This is especially important in low-resource and LMIC settings, where data quality is inconsistent, resources are uneven, and the consequences of error can be amplified.

For chronic care, the design principle is nuanced. Human oversight should be strongest where consequences are severe, ambiguity is high, or circumstances are changing quickly. Lower-risk support tasks — reminders, education, check-ins, routine follow-up — can often be safely automated within clear boundaries and under governance. The challenge is not merely to keep humans involved. It is to define which humans should be involved, for which decisions, and under what conditions.

But human-in-the-loop, while necessary, may be too narrow a frame. It implies that the machine is central and the human is inserted back into the process to supervise or correct it. For chronic disease management, a more powerful model comes from Thomas Malone’s work on collective intelligence at MIT: the idea that the most effective systems are not machines with humans attached, but groups of people and computers thinking together.

This “computer in the group” perspective is especially relevant because chronic care is already group-based. A person living with diabetes and hypertension typically relies on some combination of family members, community health workers, nurses, pharmacists, primary care clinicians, specialists, hospitals, employers, payers, and public health programmes. Yet these groups frequently operate in silos. They may not share information well, may not see the same warning signs at the same time, and often have different incentives and time horizons.

Putting the computer in the group means designing AI as a participant in coordinated human systems rather than as a parallel authority. In this model, AI helps the group notice silent risk, remember what matters, identify who needs follow-up, surface likely deterioration, and communicate more consistently across settings. Humans and institutions still own judgment, priorities, relationships, and accountability. The computer contributes pattern recognition, persistence, memory, and scale.

What this looks like in practice?

A Scenario: Chronic Care in Chiang Rai Province

Somchai is 54, lives in a rural sub-district of Chiang Rai, and was told two years ago that his blood sugar was high. He takes metformin inconsistently, has not had his kidneys checked, and does not fully understand why his doctor added a blood pressure medication at his last visit.

Today, without a connective layer, his journey is fragmented. His village health volunteer (VHV) has a paper register. The district hospital sees him twice a year. His employer — a small agricultural cooperative — offers no health engagement. His wife manages his diet but has no guidance. Nobody sees the full picture.

Now place a computer in the group. Through LINE — the app Somchai already uses daily — an AI care companion checks in weekly in Thai, asks about medication, explains why the blood pressure pill matters for his kidneys, and logs his responses. When Somchai reports persistent ankle swelling, the system flags it to a nurse coordinator, not as a diagnosis, but as a signal worth a call. His VHV receives a monthly summary. The district hospital sees his adherence trends before his next visit. His employer’s wellness programme gets aggregate, anonymised risk data for workforce planning.

No single actor replaced. Every actor better informed. The group, not the machine, delivers the care — but the computer makes the group smarter.

This framing is more than rhetorical. It changes product design and partnership strategy. It shifts the goal away from building an autonomous doctor and toward building connective intelligence for real care ecosystems.

The Regional and Rural Divide

Any serious engagement with AI in healthcare must confront the fact that chronic disease burden and access to care are unevenly distributed. Rural populations, informal workers, lower-income households, and communities distant from large referral centres often face later diagnosis, poorer continuity, and fewer options for specialist input. Even where national coverage schemes exist, physical access, workforce shortages, and fragmented navigation can make effective chronic care feel far away.

Technology can help narrow these gaps, but only if it is designed around the realities of low-bandwidth environments, intermittent connectivity, multilingual populations, and uneven health literacy. Otherwise, digital and AI tools risk primarily benefiting already-connected, better-educated, urban users — thereby widening the divide they claim to address.

People in underserved settings do not always need deep specialist AI first. Often, they need an always-available first layer of explanation, navigation, and triage: something that helps them understand risk, know whether a symptom matters, know where to go, and stay connected long enough for a human care pathway to take over. The most profound use case for AI may not be replacing advanced tertiary expertise, but making basic chronic care guidance, continuity, and escalation more available to people who currently have too little of all three.

The Care Management Challenge Across Stakeholders

The chronic care problem is shared across the ecosystem, but it appears differently depending on where one sits.

Governments and public programmes are under pressure to move from episodic treatment toward integrated, prevention-oriented NCD care, often through primary health care. Yet they must do so while managing workforce shortages, constrained budgets, variable digital infrastructure, and the difficulty of translating diagnosis into effective long-term control. Thailand’s own data illustrate this: detection and coverage of diabetes and hypertension can improve, while effective control remains stubbornly low — showing that access to care alone does not guarantee successful management.

Hospitals and clinics may have strong protocols and specialist expertise, yet still struggle with what happens between visits. Patients miss follow-up, fail to adhere to medication, misinterpret symptoms, or present late with deterioration that might have been caught earlier through better continuity and outreach.

Employers and payers increasingly bear the cost of chronic disease in working-age populations. They have incentives to invest in prevention and engagement, but often do so through siloed wellness programmes that sit apart from public systems and provider workflows — creating duplication for institutions and fragmentation for users.

At the centre of all three is the same underlying issue: no single actor owns the patient journey end to end. The lived experience of chronic disease remains fragmented even when each stakeholder performs its role reasonably well.

The Case for a Horizontal Engagement Layer

This is the opening for a horizontal solution provider. In chronic care, the missing layer is often not another specialist application or another static content portal. It is a persistent engagement and coordination layer that can sit across institutions and remain close to the person over time.

A horizontal engagement layer does several things at once. It translates guidelines and risk information into day-to-day human language. It keeps people connected between appointments. It structures common questions and symptoms so they can be understood consistently. It routes people to the right human and institutional resource when thresholds are crossed. It creates a memory of the journey that different actors can plug into. And, if designed well, it allows AI to serve the ecosystem without replacing it.

How SabaiHealth delivers this in practice?

SabaiHealth is designed as exactly this kind of horizontal care companion and triage layer — not a digital physician, not a disease-specific point solution, and not a consumer wellness app detached from care systems. It is a persistent, AI-powered engagement platform that can be configured by governments, hospitals, clinics, and payers to support the same chronic care journey from different institutional angles.

SabaiHealth: Key Capabilities and Proof Points

Delivered where people already are: GeniusCare™, SabaiHealth’s AI care companion, operates natively on LINE, WhatsApp, and Telegram — the messaging platforms that hundreds of millions across Southeast Asia use daily. No app download, no new interface, no digital literacy barrier.

Proprietary retrieval layer: SabaiHealth’s proprietary retrieval and orchestration layer blends multiple large language models to optimise for accuracy, cost, and safety — delivering intelligent, contextual health guidance at scale while keeping per-interaction costs low.

Wearable integration: Through partnerships including Huawei wearables and the ROOK SDK, GeniusCare ingests real-time biometric data — activity, sleep, heart rate — to personalise guidance and detect early signals of deterioration.

Proven engagement: In active deployments, GeniusCare achieves weekly active user rates of 23%, monthly active user rates of 35%, and an average of 12.5 messages per user per session — significantly above industry benchmarks of 7–8, indicating genuine ongoing relationships, not one-time curiosity.

Multilingual by design: Built from inception for Southeast Asia’s linguistic diversity, supporting Thai, Bahasa, and English, with architecture designed for rapid expansion.

Under this model, SabaiHealth’s value is not simply AI output. Its value is system orchestration through engagement:

• For public programmes: extending the reach of screening, follow-up, and NCD control beyond facility walls.
• For hospitals: maintaining continuity between visits, surfacing deterioration earlier, and reducing avoidable admissions.
• For clinics: structuring guidance and escalation so that scarce clinician time is used where it matters most.
• For employers and payers: improving engagement with high-risk populations and generating the data needed for evidence-based wellness investment.

Because the platform sits at the engagement layer, it can support the same person across multiple institutional contexts rather than forcing them into separate journeys for each stakeholder.

A note on data governance

Institutional partners rightly ask: where does patient data go? SabaiHealth is designed around privacy-first principles, with data residency aligned to local regulatory requirements, role-based access controls, and a clear commitment to never monetising patient data. The platform is pursuing ISO 27001 certification and is architected for interoperability with standards including HL7 FHIR, enabling integration with existing hospital and public health information systems rather than requiring institutions to replace what they already have.

What This Means for Strategy and Partnership?

The strategic implication is that no single institution can solve the chronic care problem alone. Governments need better continuity and reach. Hospitals need better visibility between visits. Employers and payers need better engagement and prevention. Patients need trustworthy companionship, navigation, and escalation. AI has the potential to support all of these aims, but only when it is embedded into collaborative delivery models rather than deployed as an isolated technological claim.

For solution providers, this means the future does not belong only to disease-specific apps or single-use AI tools. It belongs to platforms that can sit horizontally across the ecosystem, adapt to different institutional use cases, and remain grounded in human-governed care. Such platforms must be multilingual, low-friction, interoperable, and capable of functioning in both high-resource and constrained environments.

For health systems, it means AI strategy should be less about replacing clinicians and more about augmenting collective intelligence. The question is not whether AI will enter healthcare more deeply — it already has. The question is whether it will be designed as a force for fragmentation or as a force for coordination.

Conclusion

Chronic disease is not merely a set of diagnoses. It is the central coordination challenge of modern healthcare. The rise of diabetes, hypertension, chronic kidney disease, and cardiovascular disease is exposing the limits of episodic care, the costs of fragmentation, and the inequities of systems that rely too heavily on people navigating complexity alone.

Artificial intelligence can help, but only if its role is defined with precision and humility. It can support early detection, continuity, triage, engagement, and prioritisation. It cannot replace trust, accountability, clinical judgement, or the structural work of strengthening care systems.

The most useful future is not one in which humans are awkwardly inserted into machine loops after the fact. It is one in which computers are deliberately placed inside human groups — helping patients, families, providers, employers, payers, and governments work together more intelligently over time.

In that future, the role of a platform such as SabaiHealth is clear: to act as the horizontal engagement layer that turns fragmented chronic care into coordinated chronic care, and isolated encounters into sustained support.

An Invitation to Partner

SabaiHealth is actively seeking pilot partners among hospitals, clinics, employer groups, insurers, and public health programmes in Southeast Asia and beyond. Whether you are a district hospital looking to improve NCD follow-up, an employer seeking evidence-based chronic disease engagement, or a government programme working to extend care continuity into underserved communities — we would welcome a conversation about how a horizontal engagement layer could support your goals.

Contact: rohit@sabaihealth.com

Did You Sabai Today?

About the Author: Rohit Chandrasekharan Nambiar is Co-Founder & Ops Lead at SabaiHealth Pte Ltd and Member of the UNICEF Expert Advisory Panel.