Matthew Hellyar

Welcome to the New Art of Medicine — Respocare Connect AI Most of medicine is still reacting to information. A small group is starting to interact with intelligence. Not dashboards. Not scribes. An ecosystem that thinks with you — across patients, time, and uncertainty. It doesn’t replace clinicians. It removes everything that slows them down. This is what happens when behaviour becomes the product — not just technology. Now open for early access. We’ve kept it intentionally limited. If you’re a medical professional — apply now: RespocareConnectAI.com

Five questions healthcare can no longer avoid. AI is no longer sitting outside medicine asking for permission. It is already inside the workflow. Inside the note. Inside the search. Inside the risk score. Inside the clinical summary. Inside the decision support layer. So the conversation has changed. The question is no longer: “Is AI coming to healthcare?” It is: “What happens now that it is already here?” Respocare Insights exists to answer that question clearly, honestly, and without hype. This edition focuses on the five questions every healthcare professional is privately asking about AI: replacement, trust, bias, data, and liability. respocareinsights.io/weekly/five-qu…

The most dangerous AI in healthcare may not be the one that fails loudly. It may be the one that sounds right. Structured. Fluent. Confident. Clinically polished. Wrong. That is why governance matters. Not because AI is useless. Because it is powerful enough to be trusted too quickly.

@TTrimoreau Had an intersting conversation about this with an IP lawyer and my argument yes of course you can

Can you call yourself a founder if your entire product was built by AI like GPT or CLAUDE ?

Healthcare is not suffering from a lack of data. It’s drowning in it. Documents. Scans. SOAP notes. Discharge summaries. Bloods. Messages. Timelines. Contradictions. The problem is no longer collection. The problem is navigation. This is why we believe agentic AI changes everything. At Respocare Connect AI, one of the biggest lessons from active clinical trials has been this: Data should no longer be viewed as static storage. Data becomes facilitation. Data becomes memory. When structured correctly, an agentic system can retrieve, connect, prioritise, and reason across information in ways traditional software simply cannot. Not to replace clinicians. But to reduce the cognitive burden of modern medicine. The future clinical environment may not belong to the organisation with the most data. It may belong to the organisation that can safely orchestrate intelligence across that data with governance, memory, and context. That is the shift we are building toward. #AgenticAI #HealthcareAI #ClinicalAI #DigitalHealth #BuildInPublic #HealthTech #FutureOfHealthcare

Data is no longer just storage. In agentic AI, data becomes memory. And memory becomes access. That changes everything. Because the quality of an AI system is no longer determined only by the model. It is determined by how intelligently, safely, and contextually the system can access information across time. This is one of the biggest things we’re learning building Respocare Connect AI in active clinical trials. The strategy around healthcare AI has to shift. The future is not dumping documents into a chatbot. The future is governed memory architecture: • structured retrieval • patient-scoped access • longitudinal reasoning • identity-locked workflows • contextual understanding • behavioural guardrails In agentic systems, memory is what allows reasoning to happen coherently. But memory without governance becomes risk. That is why we believe the next generation of healthcare infrastructure will be designed around one core principle: The right intelligence should only access the right information at the right time for the right reason. That is where the real engineering begins. #AgenticAI #HealthcareAI #ClinicalAI #BuildInPublic #DigitalHealth #AIInfrastructure #HealthTech

Building AI for healthcare is not just about intelligence. It’s about responsibility. At Respocare Connect AI one of the biggest things we’ve learned during clinical trials is how far you actually have to go to remain HIPAA and POPIA aligned in an agentic AI environment. The AI is only one layer. The real work happens underneath: • identity-locked architecture • patient-scoped retrieval • encrypted storage • clinician approval layers • auditability • governed workflows • strict access control • behavioural guardrails People see the interface. They rarely see the infrastructure required to make it safe. Healthcare AI cannot be treated like consumer AI. The standards have to be different. #HIPAA #POPIA #HealthcareAI #AgenticAI #ClinicalAI #BuildInPublic #DigitalHealth

I’m telling you agentic AI engineering is the future. We busy also building Respocare connect AI studio designed just for this.

Ambient listening hears the conversation. Dictation captures the words. But the future of the medical AI scribe is neither. At Respocare Connect AI, we built both: • Traditional clinical dictation • Ambient AI scribing Because clinicians work differently. Some prefer structured dictation after a consult. Others want seamless ambient capture during the interaction itself. But here’s what matters most: Transcription alone is no longer enough. The real breakthrough is when the system understands what it’s hearing. Not just converting speech to text — but interpreting clinical meaning, identifying context, structuring information, and generating the right clinical note intelligently. SOAP notes. Progress notes. Discharge summaries. Referral letters. Clinical summaries. Structured. Contextual. Stored. Ready. And beyond the note itself, our agentic assistant can reason across the patient timeline, retrieve prior records, connect longitudinal history, and support clinicians with grounded clinical context. That’s the difference between: an AI recorder… and an AI clinical workflow system. We are moving from transcription to clinical intelligence. From documentation… to understanding. Respocare Connect AI Agentic Intelligence for the Art of Medicine.

@hthieblot The future of healthcare Agentic AI clinical ecosystem , clinical trails phase 3

If you are a solo founder, reply here and tell me what you are building.

One thing we will never overlook at Respocare Connect AI is education, support, and human connection. As healthcare moves into the age of AI, one concern keeps coming up: “Who is actually behind this system?” For us, that answer matters deeply. Behind Respocare Connect AI are real people. Real offices. Real support lines. A real healthcare company. Not a fly-by-night startup chasing hype. Respocare has spent nearly a decade working directly in healthcare — supporting patients, specialists, clinics, nurses, and families through oxygen therapy and medical services every single day. We understand something important: Healthcare is not just technology. It is trust. That means when clinicians use our systems, they must know there are people behind the platform who understand healthcare, urgency, responsibility, and service. As we move closer to launching Respocare Connect AI, our goal is not just to build intelligent systems. It’s to build intelligent systems backed by human support, clinical understanding, and long-term accountability. The future of healthcare may be powered by AI… but it still needs people who care on the other end of the line.

The new art of medicine

There are probably only 10 or so companies globally attempting what we’re building with Respocare Connect AI. Not AI demos. Not wrappers. Real agentic clinical infrastructure. Systems that retrieve, reason, remember context, and safely support clinicians inside real workflows. So either we’re completely crazy… or we’re very early. I believe healthcare’s future will belong to trusted agentic systems that reduce burden, restore time, and collaborate with clinicians — not replace them. South Africa deserves to build in that future too. So we keep building. Quietly. Relentlessly.

Only a handful of companies in the world are seriously attempting what we’re building with Respocare Connect AI. Not chatbot demos. Not wrapped APIs. Not “AI for healthcare” landing pages. Real agentic clinical infrastructure. Systems that retrieve. Reason. Understand longitudinal context. Operate within governance. And collaborate with clinicians safely. Sometimes I stop and ask myself: Are we crazy? Or are we simply early? Because when you spend enough time inside hospitals, with clinicians, with overloaded healthcare systems, one thing becomes obvious: Healthcare cannot scale on administration alone. The future will belong to systems that can intelligently assist clinicians across documentation, retrieval, reasoning, coordination, and decision support — while remaining transparent, auditable, and human-led. That’s the direction we’ve committed to at Respocare Connect AI. And the interesting part? There are probably only 10 or so companies globally truly building toward this level of agentic healthcare infrastructure right now. That should either terrify us… or motivate us. I choose motivation. South Africa deserves to be part of the future of healthcare AI — not just a consumer of it. So we keep building. Carefully. Publicly. Relentlessly. The future of healthcare will not be one giant AI. It will be trusted intelligence woven safely into clinical workflows. And I genuinely believe that future has already started.

@acadictive Glorious let’s !

@MatthewHellyar seems like a journey worth following ... lets connect.

Building in public. One of the biggest things we are focused on before launching Respocare Connect AI is not the interface. It’s scale. Can the system safely handle: • thousands of clinical documents • massive longitudinal patient histories • concurrent clinicians • heavy API workloads • vector retrieval at speed • agentic reasoning under pressure • real-world congestion without compromising safety That’s what we are testing relentlessly right now. Because in healthcare AI, failure doesn’t just mean downtime. It can mean broken trust. So we are stress testing everything: data architecture, retrieval discipline, workflow separation, memory handling, cost attribution, queue behaviour, auditability, and infrastructure resilience. The goal is not to launch quickly. The goal is to launch world-class. Healthcare AI deserves that level of attention. And honestly, we will not rest until we believe the system is capable of operating at that standard.

@MatthewHellyar This is the part most teams never master — the discipline to stop. In clinical AI, the real engineering isn’t the model, it’s the willingness to delay savings, document failures, and protect trust when nobody’s watching.

The hardest part of building clinical AI is being honest about what’s broken. This week at Respocare Connect AI, part of our operational audit scored 4/10. That result did not sit well with me. But instead of hiding it, softening it, or marketing around it — we documented it. Because healthcare AI cannot be built on polished demos and confidence theatre. It has to be built on discipline. On transparency. On systems that are willing to expose weaknesses before the real world does. So we stopped. We reviewed infrastructure. We investigated gaps. We tightened workflows. We resolved what needed to be resolved. One thing I’m learning building in this space: Trust is not built when everything looks perfect. Trust is built when standards remain high even when the results are uncomfortable. Healthcare deserves that level of honesty. And honest systems become safer systems.

@saen_dev Thanks saeed :) you will be pleased to know the problem has been resolved worked on this all day

Rating yourself 4/10 on operational maturity and publishing it takes more guts than any launch post. Most teams in healthtech won't even audit themselves because they're afraid of what they'll find.

@hustle_fred Indeed we have had to go the extra mile here

@MatthewHellyar makes sense, shipping fast is good but with health data it’s just way higher stakes

The hardest part of building clinical AI isn’t the AI. It’s the discipline to stop before you ship. This week we: • found a silent cost-tracking failure • identified a governance logging gap • documented operational maturity as 4/10 • delayed changes that could save $205K/year Why? Because our users are doctors. And their users are patients. Healthcare infrastructure cannot be built like startup theatre.

Startup update from the clinical AI trenches: This week our governance AI audited itself and gave itself a 7.35/10. Most startup founders would hide that. We documented it. Because if your AI can’t admit uncertainty, it probably shouldn’t be anywhere near healthcare. Also this week: • found a silent infrastructure bug • rebuilt our metering system from scratch • deployed six AI monitoring agents • delayed a change that could save us $205K/year Why delay it? Because speed is easy. Discipline is hard. One thing I’m learning building Respocare Connect AI: The future winners in AI won’t be the companies with the flashiest demos. They’ll be the ones with the strongest operational discipline when nobody is watching. Healthcare is teaching us that very quickly.