In a single quarter, every AI giant decided it could no longer stay out of healthcare. The announcements were about data and models. The screen where a clinician looks at that data and decides what to do next was almost never shown.
This piece argues that data alone — even with the best AI on top of it — does not produce outcomes. Value is the product of data and collaboration, and a zero on either side zeroes out the whole thing. Written for health-system leaders and builders evaluating the next round of AI claims.
Key Takeaways
- Q1 2026 brought ChatGPT Health, a HIPAA-ready Claude, Google's continuous Fitbit data layer, Copilot Health, Apple's FHIR R4 upgrade and a Perplexity health agent — all inside one quarter.
- Data has to become information, information has to become action, and action only matters when it becomes coordinated, team-based care.
- An insight that never reaches a person who can act on it, in time and in the right form, is a slide — not an outcome.
- Value = Data × Collaboration. Big data with no collaboration is a warehouse where the patient still falls; great collaboration with no data is a team flying blind.
- The most interesting work is the mash-up: reconciling what was documented with what was actually done, captured where care is delivered.
- One test for every announcement — ask where the team is in the picture. If the demo stops at insight, the value is theoretical.
I read another one this week. Another article on the rush to AI in healthcare. Jhanvi Shah's piece in All Health Tech on the Q1 2026 rush of AI giants into healthcare is a particularly good read.
In a single quarter, OpenAI launched ChatGPT Health, Anthropic shipped a HIPAA-ready Claude and acquired Coefficient Bio, Google tied Fitbit into a continuous data layer, Microsoft pushed Copilot Health, Apple upgraded to FHIR R4, and Perplexity entered with its own health agent. Every major AI company decided, at roughly the same moment, that they could no longer stay out of healthcare.
I get it. Across my career I have been struck by the same thing the giants are now chasing: the ever-growing mountain of healthcare data, from the literature offering best practice to the EMR giving clinical context for individual patients and populations. The mountain is overwhelming without a tool like AI to simplify it, and AI does promise faster, deeper insights than we have ever had.
But here is where I want to push back.
Data, even with the best AI on top of it, is not enough. Data has to become information, information has to become action, and action only matters when it becomes a coordinated, team-based approach to care.
That is the part the rush keeps glossing over.
The Work Happens After the Insight
An insight that does not reach a person who can act on it, in time, in the right form, is not an outcome. It is a slide. As Maria Prokhorova observed in the AHT piece, every major company launched a health product in Q1 and almost none have shown the interface — the screen where a real person looks at their data and decides what to do next. The model is the demo.
In healthcare, the model is not the demo. The workflow is. And the workflow is where collaboration lives — the bedside-to-virtual handoff, the deterioration signal that becomes a phone call, the documentation a downstream clinician can trust because the upstream team trusted it first.
Value = Data × Collaboration
Big data with no collaboration is a warehouse where the patient still falls. Great collaboration with no data is a well-meaning team flying blind. It is the product of the two that creates value. A zero on either side zeroes out the whole thing.
Most of what got announced in Q1 was a bet on the data side. The collaboration side — the part that turns insight into action and action into coordinated care — barely showed up. Collaboration is not a feature you add at the end. It is the substrate. It does not get faster because a model got bigger; it gets faster when data, the people around the patient, and the tools they share all move in the same direction at the same time.
The experts in the AHT piece keep landing in the same place. Verily's Myoung Cha notes that general models lack clinical context and personal baselines. Prashant Palepu adds that AI can pull data from six lab reports and still fail to translate it into life impact for a real patient. Both observations point at the same gap: insight stops short of action, and action stops short of a coordinated team.
The Real Opportunity: The Mash-Up
The most interesting work in healthcare AI has very little to do with bigger models. It is the mash-up.
Consider one example in telehealth. On one side, what is documented — the note, the order, the EMR entry. On the other, with virtual care, what is actually done — captured on video, in real time, in the patient's room. These two streams have lived in parallel universes. The note is what we say happened. The video is what happened.
What Shows Up When You Bring Them Together
- A continuous feedback loop on care delivery.
- An honest answer to telehealth's actual ROI.
- A coaching layer for teams.
- A safety layer — mismatches between documentation and practice are where harm tends to hide.
That is Data × Collaboration, captured where care is actually delivered, by the actual team delivering it.
The Test for What Comes Next
As you read the next round of AI announcements, I would offer one test: for every claim about data and AI, ask where the team is in the picture. If the team is missing — if the demo stops at insight — the value is theoretical.
Both sides at once
The companies that will move outcomes are working on both sides at once. Closing the loop between what was documented and what was done. Putting AI inside the team rather than next to it. Data times collaboration. That is where the value is.
FAQ
What does “Value = Data × Collaboration” actually mean?
It means value in healthcare is a product, not a sum. Data and AI supply the insight; collaboration turns that insight into action and action into coordinated care. Because the two multiply, a zero on either side zeroes out the result — which is why a large data advantage with no collaboration layer produces very little at the bedside.
Isn't better AI enough to improve outcomes?
Better models produce better insight, and insight is not an outcome. An insight that does not reach a person who can act on it, in time and in the right form, is a slide. Outcomes move when the insight lands inside a workflow that a team already shares.
What was actually announced in Q1 2026?
OpenAI launched ChatGPT Health, Anthropic shipped a HIPAA-ready Claude and acquired Coefficient Bio, Google tied Fitbit into a continuous data layer, Microsoft pushed Copilot Health, Apple upgraded to FHIR R4, and Perplexity entered with its own health agent. Nearly all of it was a bet on the data side.
What is the “mash-up” in telehealth?
Reconciling what was documented with what was actually done. The note, order and EMR entry on one side; the video record of care delivered in the patient's room on the other. Bringing the two streams together yields a feedback loop on care delivery, an honest ROI figure, a coaching layer for teams, and a safety layer.
Why is documentation-versus-practice a safety issue?
Because mismatches between the two are where harm tends to hide. If the chart and the delivered care diverge and no one reconciles them, the downstream clinician is acting on a picture that was never true.
How should I evaluate the next AI announcement?
Ask one question: where is the team in the picture? If the demo stops at insight — no interface, no handoff, no coordinated action — the value is theoretical.
Sources
- Jhanvi Shah, All Health Tech — Q1 2026 analysis of AI giants entering healthcare, including commentary from Maria Prokhorova, Myoung Cha (Verily) and Prashant Palepu.
- Company announcements, Q1 2026: OpenAI, Anthropic, Google, Microsoft, Apple, Perplexity.