Four ways to buy this, and when each one is right
Most of this decision is made before anyone looks at a feature list. It comes down to how many locations you run, how many systems they run on, and who is accountable for the number at the end.
| Criterion | Practice management SaaS PMS, EHR, engagement | Point AI tool Receptionist, scribe | Model + engineers Custom build | AI Operating System Samara |
|---|---|---|---|---|
| What you are buying | A tool your staff operate | One task, automated | Capacity to build | Outcomes, produced and traced |
| Judged on | Features and UI | How well it does its one job | Scope of what is possible | Results, and whether they hold up |
| Who does the work | Your staff | The tool, for one step | The vendor's engineers | Agents, under human approval |
| Spans multiple systems of recordThe question that decides everything at scale | No, one at a time | No | If you pay to build it | Yes, by design |
| Requires replacing your EHR or PMS | Often, eventually | No | Depends on the build | No |
| One metric definition across every location | Only on migrated sites | No | Built once, maintained by you | Yes |
| Absorbs the next acquisition | A new migration | A new install | A new project | A new connector |
| Where the human sits | Doing the work | Supervising one step | Writing the requirements | Reviewing, approving, denying |
| What you own at the end | A subscription | A subscription | A codebase and its maintenance | An operating layer and its evidence |
| Typical pricing basis | Per seat, per feature tier | Per seat or per interaction | Time and materials | Modules used, moving toward outcomes |
Three of these four columns are the right answer for someone. The guide below says who.
Which one is you
We would rather you buy the right thing than buy from us. Three of these four are not us.
One practice, one system, one location
Practice management SaaSIf everything already lives in one place, a good PMS will serve you better and cost less than anything on this page. Buy the category leader and move on.
The phones are the problem, and only the phones
A point AI toolA well-built single-purpose tool will fix a single-purpose problem faster and cheaper than a platform. The limit arrives when the second problem does.
Something genuinely novel, and engineers to spare
Model plus a build teamIf the requirement is unique to you and you can carry the maintenance, building it is legitimate. Budget for the second year, not just the first.
Multiple locations, more than one system of record, and no single answer to how the group is performing
An AI Operating SystemThis is the case that no application can serve, because the question is about the group rather than any one site. It is the only situation where we are the right call.
You use five percent of it
Decades of feature stacking produced software that does everything and asks you to navigate all of it. Most practices need a small fraction of what they are paying for, and they pay for the rest in training time and daily friction.
The burden of learning complex software used to sit with the user. That is the part that moves.
And switching was made expensive on purpose
Certification programmes, proprietary formats and training investments have functioned as retention mechanisms in this industry for twenty-five years. The cost of leaving, not the value of staying, is what keeps many contracts renewing.
An operating layer that sits above your systems does not need you locked in, because it never asked you to leave them in the first place.
For sponsors and platform leadership
The comparison changes if you are holding the asset
An operator is comparing tools. A sponsor is comparing what the platform looks like to whoever buys it next. Those are different questions, and only one of them is answered by a feature list.
Acquire
What are we actually inheriting?
Map the systems and the operating variance across every site before it becomes a surprise.
Operate
How do we make forty locations behave like one?
Connect what is there, standardize the workflows, automate the repeatable work.
Exit
How do we prove any of it?
Per-location evidence that the operating model transfers, rather than an assertion that it does.
Standardization work that is instrumented from the start produces evidence at exit. The same work, uninstrumented, produces a line in the operating expense column and nothing to show for it.
Still not sure which column you are in?
Tell us how many locations and how many systems. That answers it faster than a demo.