A surprising amount of a therapy practice can live inside people's heads.
Your admin knows that one insurance plan has to be handled differently from the others. Your biller knows what to check when a payment doesn't look right. Someone remembers that a particular clinician is technically available on Fridays but doesn't really want new clients then. You know which employee needs a little more follow-up, which recurring expense hits at the end of the month, and what needs to happen before payroll runs.
None of this feels like a system problem because the practice keeps working.
Until the person who knows something isn't there. Then a routine question becomes a Slack message, a text, or a phone call. Someone searches through an old email thread. A task sits unfinished because nobody is sure what happens next.
The information existed. The practice just didn't own it. A person did.
Human Memory Works Better Than It Should
There is a reason practices end up operating this way. When a business is small, memory is incredibly efficient.
If I have three clinicians, I probably know their schedules, specialties, insurance panels, and current availability. Writing all of that down somewhere and maintaining it might actually take longer than remembering it. The same thing happens with administrative processes. One person learns how credentialing works. Someone else figures out payroll. The owner knows how to handle unusual client situations.
For a while, this works remarkably well. Then the practice grows.
Now there are ten clinicians instead of three. Availability changes every week. Different clinicians take different insurance plans. Someone is licensed in another state. A new admin is hired. Policies change. Exceptions accumulate. The information that used to be easy for one person to remember becomes an increasingly complicated web of institutional knowledge.
But because the transition happens gradually, the practice often keeps using the same basic system: ask the person who knows. That works until the person who knows isn't available.
Look for the Person Everyone Asks
One of the simplest ways to find institutional-memory risk is to notice who has become the answer to everything.
Most practices have at least one person like this. It may be the owner. Often it is an experienced admin or biller who has been there long enough to know how everything actually works. That person is usually extremely valuable — and can also become one of the practice's biggest operational vulnerabilities.
If someone has to ask the same person how to handle a process, find a document, interpret an exception, or remember what comes next, the practice is depending on that person's memory. This isn't a criticism of the employee. Usually it means they became good enough at their job that the organization gradually built itself around them.
I've seen this happen in my own businesses. A capable person solves problems, remembers the exceptions, and gets faster at handling them. Because they're good at it, there's very little pressure to build a better system — until they take a vacation, change roles, get sick, or eventually leave. Then you discover how much of the business they were carrying around in their head.
Not Everything Needs to Become an SOP
The obvious solution is to document everything. That's not quite the right answer.
Some information belongs in an SOP. Some belongs in a database, checklist, calendar, task system, or clinician directory. Some simply needs to be captured at the moment a decision is made. The format matters less than whether the information is available where the work happens.
If clinician availability changes frequently, a six-page SOP describing availability is the wrong tool. You need a current source that tells the person handling an inquiry who is actually accepting clients. If payroll has a consistent sequence of steps and checks, that may deserve a documented procedure. If a credentialing deadline occurs three months from now, it belongs in a system that will surface it when it needs attention — not in someone's memory and not buried in meeting notes.
The mistake is treating documentation as the solution to every knowledge problem. What you actually need is a reliable place for important information to live outside a person's head.
Pay Particular Attention to Exceptions
Routine work is usually not where institutional memory causes the most trouble. Exceptions are.
Most employees can learn the normal process fairly quickly. The knowledge gap appears when something doesn't fit. What do we do when a prospective client wants a clinician who is full? What happens when insurance pays differently than expected? Who handles a complaint about a clinician? What if someone terminates employment while they still have active clients?
The experienced person often knows the answer because they've encountered it before. Everyone else asks them.
You cannot document every unusual situation a practice will ever encounter, and trying to would create an unusable manual. But recurring exceptions are different. If the same question has come up three or four times, the practice has learned something, and that knowledge should probably become part of the system — whether that means updating an SOP, adding a field to a directory, creating a decision rule, or fixing the workflow so the exception stops happening.
The point is to let the organization learn, rather than requiring the same person to remember the answer indefinitely.
Build Systems Around Retrieval, Not Storage
Practices usually don't have a shortage of stored information. They have Google Drive, email, the EHR, spreadsheets, shared folders, meeting notes, Slack. Maybe an operations manual.
The problem is finding the right information when someone actually needs it.
A beautifully organized folder nobody thinks to open is not much better than information living in someone's head. When I think about an operational system, I care a lot about retrieval. Where would a reasonable person look for this information while doing the task? Can they find it quickly? Is it obvious that what they found is current? If something changes, does the person responsible know what needs to be updated?
This is also one of the areas where AI may eventually be useful — not as the source of truth, but as a better way to retrieve information from sources the practice already maintains. The underlying requirement doesn't change, though. The information has to exist somewhere reliable first. AI cannot retrieve institutional knowledge that was never captured.
Get the Important Things Out of People's Heads
You do not need to eliminate human memory from your practice. People will always know things that aren't documented, and experienced employees will always develop judgment that can't be reduced to a procedure. That's valuable.
The risk is when ordinary operations require that specific person to be available.
A useful place to start is a simple question: if someone on your team unexpectedly disappeared for a month, what knowledge would disappear with them?
Don't try to document their entire job. Find the information that would prevent other competent people from continuing the work. Capture the recurring decisions. Identify the deadlines they are keeping track of. Make current information easy to retrieve. Put ongoing processes somewhere they can be maintained instead of merely stored.
Then do the same thing with the next critical role.
The goal isn't to build a practice where nobody needs to think. It's to build one where people can use their judgment on problems that actually require it — instead of spending their time trying to remember, or find someone who remembers, how the practice works.