Research pillar · Engineering Care Delivery
The Competitor May Look Like Your Customer
A regulated technology business may find that its next competitor is not another vendor but the practice it has been selling to, rebuilt around AI as an operating company.
The question
What happens to a regulated technology business when its own customer rebuilds itself around AI native care delivery?
Active · since 2026 · Engineering Care Delivery
The rebuild signal
Established digital health and medical technology companies are noticing the same capability shift that new entrants are acting on. Some are asking whether their business should be taken apart and reassembled around AI, and whether the resulting company is a care delivery organization rather than a software or solution vendor. This is an observed discussion, not a settled trend, and very few completed rebuilds can yet be inspected.
A second observation sits beside it. Founders who already built and ran mature regulated technology businesses are returning to operating roles at an unusual moment in the technology. Their return is a signal about perceived opportunity. It is not evidence that the rebuild works.
The aggressive version is to become the practice
The sharper examples come from companies with no legacy product to defend. Radley describes itself as an AI native radiology practice. By its own account the founders are not only selling AI software to radiologists. They are acquiring a radiology practice and intend to rebuild the clinical and operational stack with agents from the ground up. Their stated argument is that point solutions cannot repair an operation built around obsolete systems and processes.
Everything in that paragraph is a company claim and should be read that way. The evidence that would support it is specific: read volumes and turnaround times before and after the rebuild, radiologist hours per study, error and addendum rates, payer mix and realized rate per study, staffing cost per read, and whether any of it holds once the initial capital and founder attention move on.
Category boundaries stop holding
The threat an incumbent should sit with is not a better application. It is that the competitor stops belonging to a category. A practice rebuilt this way can compete at the same time with the radiology software company, the services company, the staffing company, and the practice next door. It reaches the patient, holds the clinical authority, bills the payer, and owns the workflow, so it can undercut on any one of those fronts while earning on the others.
That is worse for an incumbent than a rival vendor for a practical reason. A vendor competitor buys software, sits in procurement, and can be met with a product response. An operating competitor buys nothing from you, never enters your sales process, and improves on a different cost curve. Whether the economics actually work at scale is unproven, and capital intensity, clinical liability, staffing, and reimbursement all remain open questions.
Read it through the rest of this area
The pattern is a specific instance of arguments already under examination here. Care Delivery as the Product asks whether value migrates from technical creation toward patient access, clinical authority, workflow, reimbursement, and outcomes. An AI native practice is that hypothesis expressed as a company. Market and System Conditioning asks who must say yes before routine use is possible. An operator that owns the practice removes several of those parties from its own path while remaining exposed to payers and regulators.
Evidence, Economics, and Rights supplies the discipline. The five tracks stay separate. A working agent stack is scientific and technical evidence. A rebuilt practice with better turnaround is operating evidence. Neither establishes that the model earns a durable return, that reimbursement holds, or that the advantage survives a competitor copying the same approach with more capital.
What to watch, and what would change the reading
The useful posture is to watch for observable events rather than announcements. Keep the record separated into what was observed, what a company claimed, what follows by inference, and what remains a hypothesis.
The reading should weaken if rebuilt operations show no measurable advantage over well run conventional practices, if payer behavior penalizes the model, if clinical liability and professional rules limit how far agents can carry the work, or if acquired practices revert to conventional staffing once growth demands it. It should strengthen if independent operating data appears, if incumbents begin acquiring practices themselves, or if payment moves toward parties that hold outcome accountability.
| Signal | What it would show | What it does not settle |
|---|---|---|
| A technology company acquires a clinical practice | The model is being attempted with real capital | That the rebuilt operation performs better |
| Published throughput or cost per unit of care | An operating result under specific conditions | Transferability to other sites or payers |
| Payer contracts written with an AI native operator | Payment recognizes the delivery form | Durable rates or margin |
| Incumbent vendors buying delivery assets | Category boundaries are moving | Which side of the boundary wins |
| Clinical or professional rules constraining agent work | The practical ceiling on substitution | That the model fails, only where it stops |
Put this to work
Care delivery readiness map
Use the question “What happens to a regulated technology business when its own customer rebuilds itself around AI native care delivery?” on a real case and produce a record another person can challenge.
- For
- Clinical, operational, product, reimbursement, and implementation leaders.
- What you keep
- A care delivery readiness map you can review, revise, and send.
- What counts as sound
- Names a patient or care-team result
- Identifies every required handoff
- Separates status claims from evidence
- Names the party carrying cost and risk
- Includes a smallest live test
Clinical, regulatory, privacy, and release decisions remain with qualified people.
Nothing entered here is stored or sent. Review the prompt before sharing confidential, personal, patient, or privileged information.
Review the prompt
You can leave any field blank. The prompt will mark it as not provided.
If the record survives your review, send the question, evidence, unknowns, and requested next step.
