What AGI Really Means for Healthcare, and the Intelligent Operating System Behind It

Our first research paper described what healthcare becomes when AGI (Artificial General Intelligence aka super intelligence) takes over the administrative work that runs it. This article steps back to answer a more basic question: what is AGI, what does it mean for the people who pay for and manage healthcare, and what has to exist underneath it first. We call that underlying system an intelligent operating system, the part almost no one talks about, even though it decides everything.

First, What Are We Actually Talking About?

Picture Shelly, a benefits director at a mid-sized employer. Every year Shelly pulls claims from one system, pharmacy from another, and eligibility from a third, then builds a spreadsheet that is already out of date by the time anyone reads it. Formulary decisions get made in the fall using data from the spring. Keep Shelly in mind, because everything that follows is really about whether next year looks any different. And there's a principle underneath all of it that we hold ourselves to: AI-Assisted, Human-Verified. The system carries the work; a person stands behind the judgment that matters which counts for everything when the data is as sensitive as someone's health.

The AI most people have used so far is good at one thing at a time: answer a question, summarize a document, flag an unusual claim. Think of these as individual building blocks. Each is useful on its own, but each still needs a person to decide what to do next.

AGI, or artificial general intelligence, is what you get when those blocks are assembled into a working, self-improving system. If today’s AI is the box of Lego pieces, AGI is what happens when those pieces begin assembling themselves; a system that keeps learning, adapting, and getting smarter with every decision it makes. Unlike a finished Lego model, an AGI system is never truly “complete.” It is always evolving. It can reason, plan, and act on its own across a whole job, at the level of a highly trained expert. The simplest way to say it: today’s AI helps a person do a task faster, while AGI can run the task itself and bring in a person only when judgment is genuinely needed. Leading labs expect this shift within one to ten years, soon enough that the organizations paying for healthcare should understand it now.

Why the Foundation Comes First

There is a comfortable version of this story in which AGI simply shows up as a new feature, gets switched on inside the tools healthcare already uses, and quietly makes them better. That version is wrong. Intelligence is useless if it cannot see the information it needs. A system that cannot look at eligibility, claims, pharmacy, and clinical details in the same place, at the same time, is an expensive engine with nothing connected to it. The hard part is always the system underneath that supports it.

That is the uncomfortable truth most organizations are still avoiding. The intelligence is arriving faster than the connected information system that would let anyone actually use it. The gap between those two timelines is where the next competitive divide opens, and it is opening now.

The Data Problem Hiding Underneath

For any of this to work, it needs clean, connected, up-to-the-minute information. That sounds obvious until you see how employer healthcare stores it. Claims live in one system, pharmacy in another, eligibility in a third, and clinical information is scattered across portals that were never built to talk to each other. The information exists; it just does not exist in one place where a system could act on it while a decision is still being made.

Even connected information does not settle everything instantly. When someone gets care, the provider does not bill right away. Coding, administrative processing, and billing cycles all introduce a delay, so claims routinely arrive weeks or months after the service. Insurers account for the care that has happened but has not yet come through the pipeline with an estimate called IBNR, for "incurred but not reported." This matters for real time decision making because a picture built only on claims already received always understates what is truly owed; part of the cost is still in transit. A real operating system has to treat what is already known and what is still developing as two different things, and reason about both, rather than mistaking a partial claims feed for the full financial reality.

Fixing this is not a small upgrade. An organization that wants to use AGI needs all of that information brought together into one connected picture, not another dashboard on top of the same disconnected sources. Everything built on top inherits its limits: feed even the smartest system fragmented, half-complete information and you get fragmented, half-complete decisions, just produced faster and with more confidence.

The Intelligent Operating System

The simplest way to picture this is as an operating system, the way the software on your phone is one. You do not think about what happens when you tap the screen; the system handles the details so you get the result you wanted. Healthcare needs the same thing. It comes together in three parts, each depending on the one before it, and together they describe what it takes to move from software that reports what already happened to a system that can actually act.

Part One: One Connected Source of Information

This pulls claims, pharmacy, eligibility, and increasingly genetic information together into one place that updates in real time, not once a night. A file that updates only overnight is already out of date by the time anyone reads it, so it can only explain what already went wrong. Everything else depends on getting this part right first.

Part Two: Clear Rules for What Runs on Its Own

Raw access to powerful AI is not a system on its own. This part sets the rules: what the AI is allowed to do, how far it can go, and where it has to stop and hand off to a person. In practice, routine, well-understood tasks, like eligibility checks, formulary checks, and standard approvals, get handled automatically in seconds, while anything complicated or high-stakes is passed to a person with the full picture already attached. A human stays in the middle and owns the final judgment on the decisions that carry weight.

Part Three: Always On Visibility

This is where the change becomes visible to the people running a health plan. Instead of reviewing what happened once a year, they get continuous, real-time visibility into where money is going, how the drug list is performing, and what is happening across their members. It is the difference between a plan that gets a yearly check-up and one that watches and corrects itself as things unfold.

Vendor sprawl is one place this matters. Over time, employers tend to add point solutions faster than they retire them, and usage of any given one can quietly fade long before anyone notices at renewal. This is exactly the kind of drift that only shows up when someone is watching continuously rather than once a year, and it is a natural example of what always-on visibility is for: keeping the picture current enough that questions like "is this still being used, and is it still worth it?" can be asked while there is still time to act on the answer.

None of these three parts is optional. Powerful AI with no connected information has nothing to work with. Clean information with no rules around it is capability no one can safely turn on. The intelligent operating system is all four parts working together, or it is not really an operating system at all.

Four Forces Shaping What Gets Built

This intelligent operating system is not being built in a vacuum. Four bigger forces are now bearing down on it, and each one raises the stakes for any organization treating its information foundation as an afterthought. We will keep these short here, since some of them deserve a post of their own, but they are worth knowing about.

First, the quantum threat, sometimes called “Q-Day.” All of this rests on information that has to stay private for decades. Q-Day is the point at which quantum computers become powerful enough to break today’s encryption. Attackers are already collecting encrypted health data to unlock later, an approach known as “harvest now, decrypt later,” and healthcare is especially exposed because records stay sensitive for a lifetime. NIST finalized its first post-quantum cryptography standards in 2024, so this protection has to be built in from the start, aligned to those standards, not bolted on later.

Second, the cost and control of computing power. The powerful AI in all of this runs on enormous amounts of computing power, and that is getting harder to supply as data centers run into limits on electricity, water, and cooling. Those limits are already pushing compute in new directions. Some builders are trying to move it off the ground entirely: Starcloud, backed by SpaceX launches and Nvidia hardware, put a data-center-class GPU in orbit in late 2025 and is scaling toward larger satellites that run on solar power and shed heat by radiating it into space, and SpaceX itself has filed plans for large satellite constellations aimed at compute. These orbital efforts are early and unproven at scale, and most observers expect them to stay a small share of total capacity for years. The point is not that healthcare's compute will run in space soon, it is that the pressure on power and cooling is now severe enough to send serious money looking for exits, which only sharpens the underlying lesson.

The computing power is consolidating into a small number of giant providers. The lesson for healthcare is that it is becoming a strategic dependency controlled by a few hands, so today’s choices should keep the freedom to move between providers.

Third, the safety of the AI itself. As these systems get more capable, the risks they carry grow too, so the choice of which AI to use is not neutral. The leading labs are formalizing this: Anthropic, for example, ties stronger safeguards to more capable systems and holds its most powerful tier back from open public release. For healthcare, choosing an AI is partly a question of how seriously it is secured, and how much trust can be placed in the controls around it, because a system that can take real actions inside benefit and clinical decisions is one whose mistakes become real-world events.

Fourth, the rules. None of this gets to set its own rules; a fast-moving and unsettled set of regulations does. State AI laws, guidance from agencies like the FDA and CMS, and the EU AI Act all create obligations for anyone running automated systems in healthcare, and that picture keeps shifting. The details will keep moving, but the direction will not: being able to explain decisions, keep humans in the loop, and govern these systems with transparent, trustworthy controls over time is becoming a basic requirement.

How Forsure Fits Into This

This is the intelligent operating system Forsure is being built around, and the order of the four parts is deliberate. The connected foundation comes first.

For the connected foundation, Forsure pulls everything into one place the intelligence layer can actually use. Claims, pharmacy, eligibility, and member health markers are the starting point, but the picture has to include the rest of the systems an employer is already running: the HR and business tools where people and coverage live (Workday, ADP, ServiceNow, Zendesk), the point solutions sitting on top (telehealth, MSK, and so on), and the clinical systems where the care itself gets recorded (EMRs). The intelligence is only as good as what it can see, so the foundation has to reach across all of it. This is the part everything else depends on, and the part most organizations have not built.

For choosing the right brain and setting the rules, routine decisions, like formulary checks, eligibility checks, and standard approvals, happen automatically in seconds, while the bigger calls, like covering a new drug or changing a plan, go to a person with the full picture already attached. Routine work moves at machine speed; the decisions that carry weight stay with a human in the middle, who owns the final judgment.

For always-on visibility, Forsure helps employers move from once-a-year reviews to continuous, real-time optimization, paired with hands-on guidance on plan design, drug strategy, and pharmacy negotiations. The technology surfaces the decision; people frame the strategy around it.

The intelligent operating system also creates an entirely new distribution model for healthcare and insurance. The traditional consulting and brokerage model has done a lot well, but bending the cost curve has never really been what it was built for. Forsure changes the economics of that layer by doing the heavy operational and clinical work through the operating system itself, offering employers a modern alternative to the way benefits have traditionally been brokered, and giving insurance companies a partner built for where the industry is heading. Forsure is operational, clinical, and distributional, all at once.

Put together this way, the four parts stop being a diagram and become something real: a connected foundation, the right brain and clear rules, always-on visibility, and a new distribution model, with the four bigger forces designed in from the start. For Shelly, it means next year’s review need not look like last year’s.

‍