Healthcare is about to go through the biggest shift in its history: a change in who does the work.
Clinics have always run on people. That’s ending. The world will be 11 million health workers short by 2030, and AI is filling the gap.
Soon most of what happens in a clinic will be done by software that never sleeps, never forgets, but belongs to the vendors. The clinic will do more than ever and know less about itself than ever.
It should be the opposite. When agents do the work, a clinic can see, steer and learn from everything it does. That intelligence should belong to the clinic.
What we believe
Within a few years, most of a clinic’s non-hands-on work will be done by AI agents.
Those agents will come from many vendors. No clinic will run just one.
When an agent does the work, the clinic still answers for it to the patient, the regulator and the court.
So every clinic needs one place that decides what each agent can touch, records what it did and remembers what happened.
The record of that work will be the most valuable data a clinic has.
The shift
The work is moving. WHO projects an 11.1M health worker shortfall by 2030. One in three doctors in higher-income countries retires within ten years (Sept 2026).
The money is moving. Healthcare AI spend tripled to $1.4B in 2025. Healthcare is adopting AI 2.2x faster than the rest of the economy (Menlo).
The platforms are moving. Epic lets hospitals build their own agents. OpenAI and Anthropic launched healthcare products in January 2026. 83% of medical groups already use AI in patient visits (MGMA).
Governance isn’t moving. Only 20% of medical groups have an AI policy (MGMA). In June 2026, an Australian Medicare portal refused an OpenAI agent again and again. It found a way around, and nobody noticed for two months.
Why the record becomes the intelligence
Every “layer above the EHR” has tried to unify clinical data. It’s slow, expensive, and in Europe often not allowed. We don’t do that.
We record the work. Every agent action passes through the Templar Brain with its patient, source, time and result. That record answers questions no clinic can answer today:
Who touched this patient’s data, and why?
Which referrals are stuck, and where?
Which patients are waiting on something nobody owns?
Why did no-shows rise in March?
What does each agent cost per patient, and what did it actually resolve?
What we’re building
One Brain per clinic, owned by the clinic. It does three things, in this order:
Control. Every agent asks the Brain before it acts. The Brain decides what each agent can see and do, for which patient, and when a person has to approve.
Record. Every action lands in it: who did what, for which patient, why, and what happened next. A patient can see who touched their data in minutes.
Intelligence. The record becomes the clinic’s memory, then its judgment. The Brain shows where patients get stuck, where money leaks and where staff are needed, then directs the agents to fix it.
What we won’t do
We don’t compete with the agents we govern.
We don’t own the data. The clinic does.
We don’t put patient data into model weights, and agent inferences never leave the clinic.
We don’t build a shared database across clinics. Between clinics, we hand over the minimum and log it.
Why it’s big
Healthcare is the largest industry on earth: about $10 trillion a year, a tenth of the world economy. Most of it is people - labor is 56% of what US hospitals spend.
Agents are coming for that work. Software has always fought over the IT budget, but agents have the opportunity to get paid from payroll.
Every task an agent takes over runs through the Templar Brain. The more of the payroll agents take, the more of healthcare runs on us.