The Agreements Nobody Wrote Down: What Your SOPs Miss

Most processes have two versions. One lives in the documentation. The other is the one people actually follow. That second version is shaped by agreements made in meetings, in corridors and over the phone. Nobody wrote those agreements down. Often nobody remembers they are agreements at all. They just feel like the way things are done here.

This is the part of a role that disappears first when someone leaves. It is also the part a new hire cannot find by reading.

What gets written down and what doesn't

Ask someone to document their job and they will write down what they can easily explain. The steps, the systems and the deadlines. That is useful. It is also the part a capable newcomer could work out in a week.

What they leave out is not hidden on purpose. They leave it out because they no longer see it. The philosopher Michael Polanyi described this in 1966 in The Tacit Dimension: "we can know more than we can tell." Experienced people use knowledge they have stopped noticing.

So the instruction "write down what you know" has a bias built in. It collects the visible part of the job. It misses the part that made the person hard to replace.

The agreements layer

Tacit skill gets most of the attention in this debate. Think of the engineer who hears a fault before the alarm goes off. That knowledge is real, but it is hard to transfer in any format.

A second layer gets far less attention and is much easier to capture. It consists of explicit rules that were agreed but never recorded. A few illustrative examples:

• Finance sees every quote above a certain amount before it goes out.

• Approval formally goes to the department head. In practice it goes to Sandra, and he signs what she has checked.

• Releases never go out on the last day of the month, because of the billing run.

• One particular customer gets a phone call before an invoice arrives.

None of these are instincts. Each is a clear rule that someone could state in one sentence. It was settled in a conversation. It worked. Nobody went back to update the process document.

These are undocumented agreements. They are concrete and specific to your company. Nobody outside it can tell you what they are.

Incomplete is the wrong word

The usual complaint is that documentation is incomplete. That undersells the problem.

An incomplete document has visible gaps. A new hire who hits a gap stops and asks. The bigger risk is a document that looks complete and is wrong. A blank space on a map makes you ask for directions. A road on the map that no longer exists makes you drive.

Picture a new hire who follows the approval process as written and sends the quote to the department head. Nothing breaks that day. The document was followed. But the quote went out without the check everyone assumed would happen. Nobody finds out until something goes wrong. It is why role-specific onboarding has to cover the real approval routes as well as the formal ones.

Work as imagined and work as done

Safety researchers have a name for this gap. Erik Hollnagel, one of the founders of resilience engineering, separates work-as-imagined from work-as-done. In a 2015 white paper with Robert Wears and Jeffrey Braithwaite, work-as-imagined describes what should happen under normal conditions. Work-as-done describes what actually happens as work unfolds in a complex setting.

Their point is that the two always differ, and that the difference is mostly useful. People adjust procedures to conditions the procedure never foresaw. Those adjustments are a large part of why things go right.

Labour disputes offer a plain demonstration. In a work-to-rule action, employees follow every official rule exactly and do nothing more. Output drops. If following the documented process to the letter slows the work down, the documented process was never how the work got done.

Undocumented agreements are a large share of work-as-done. They are adjustments that became routine.

Where writing everything down is the right answer

This is not an argument against documentation. In some work, writing everything down and following it exactly is the right design. Aviation, pharmaceutical production and regulated manufacturing are the clearest cases. The work is stable and repeats. A deviation can be catastrophic. There the effort goes into keeping procedure and practice identical, and every informal change gets reported.

Most roles are not like that. Account management, operations, finance, product and team leadership involve changing situations and constant judgment. In those roles, work-as-done keeps drifting away from the document. So the useful question is how to see the drift.

Why people can't document their own agreements

The obvious next step runs into the same problem as before. If the goal is to capture the agreements, you would ask the person who holds them to write them down. But the agreements feel like normal work to that person. Sandra does not think of herself as an approval step. She thinks of herself as someone who reads the quotes.

So the failure sits in self-documentation. A person cannot see what is missing from their own description. From the inside, nothing is missing.

Someone who does not already know has to do the noticing. A new colleague does this without trying, which is why they ask so many questions in their first weeks. A structured interview does it on purpose. A good interviewer asks about exceptions, detours and the people who get called before something happens. Those questions surface agreements the person would never have listed. We have collected them in our guide to knowledge transfer interview questions.

What happens when one side of the agreement leaves

An agreement has two parties. While both work at the company, it holds. When one of them leaves, it becomes hard to show it ever existed.

The person who stayed may remember it differently or not at all. A new manager has no reason to honour a rule nobody can show them. The agreement lapses without anyone deciding it should. The risk it was managing returns.

This is one reason handovers fail more often than companies expect. When we asked professionals how well their handovers work, fewer than one in five called them effective. The cost of a resignation is rarely just the open workload. Often it is a check that quietly stopped happening.

A test you can run this week

You don't need a tool to see how large this layer is in your company. Pick one documented process that matters, such as quoting, client onboarding or approving spend. Ask someone new to follow it exactly as written. They change nothing and fill no gaps from common sense.

Then count how often a colleague has to step in. "That goes to Sandra first." "We stopped doing it that way last year." "You have to call them before you send that."

Each intervention is an undocumented agreement. Write it down with who made it and why. After one run you have a list that no procedure review would have produced.

What to capture instead

The output is still a document. What changes is the input. Instead of asking people to describe their job, ask about the moments where the job departs from its description:

• Who do you check with before this goes out, even though the process doesn't say so?

• What do you do differently for particular customers, suppliers or colleagues?

• Which written rule do you skip, and what do you do instead?

• What would go wrong if a new person followed the process exactly?

• Who agreed to this, and when?

The answers belong next to the process, where the next person can find them. For more on why documents alone fall short, read how AI-powered knowledge transfer goes beyond traditional documentation.

Want to find the agreements in your own team? See how Rinto captures what documentation misses, on one of your own roles.

Sources

• Polanyi, M., The Tacit Dimension, University of Chicago Press, 1966.

• Hollnagel, E., Wears, R. L. and Braithwaite, J., From Safety-I to Safety-II: A White Paper, The Resilient Health Care Net, 2015.

• Work-to-rule, Wikipedia.

• Rinto survey of professionals on handover practice.