In a company of forty people, there is a layer of work that has no name and no chair. It is the chasing of the report that did not arrive, the translation of what sales promised into what operations can deliver, the noticing that two teams are about to collide before they do. In a larger company this belongs to a middle manager. In a company of forty it belongs to no one, which means it belongs to the founder, or to whoever was standing closest when it first appeared.
The work is real, but it does not add up to a job
The reason no one hires for this layer is that the work is fractional and scattered. It might be thirty percent of a coordinator, twenty percent of a reviewer, and fifteen percent of someone who can read a contract. Those fractions do not live near each other, and they do not add up to a coherent position you could write a description for.
A hire is indivisible. You cannot bring on thirty percent of a person. So a founder who feels this layer pulling at the edges of the week has two options, and both of them are bad. The first is to hire ahead of the need and carry a full salary for a role that is two-thirds idle. For a bootstrapped company that option is usually off the table before it is spoken aloud. The second is to take the fractions and pile them onto someone already there, or onto themselves.
A person in this layer feels the ground move
When the fractions land on a real person, something happens that is easy to miss. The job they were hired for begins to drift. A person hired to manage vendor relationships finds themselves arbitrating between two departments on Tuesday and rebuilding a broken handoff on Thursday, and none of it was in the description they accepted. The description has not been rewritten. It has just quietly stopped being true.
Most people can feel this even when they cannot articulate it. The work expands one week and contracts the next. The thing they are accountable for keeps moving, and they are reviewed against a role they no longer occupy. This is the predictable result of asking one human to hold a shape that was never a shape.
There is a cost to keeping someone in a seat like that, and it shows up on no ledger. It shows up in the slow erosion of the sense that one's work has edges, that effort connects to a known result, that the role one accepted is the role one has. A person can absorb that for a long time, and usually does, because the alternative is leaving and because the founder is clearly working harder than anyone. The absorbing is the problem. It hides the structural mistake by paying for it with someone's clarity.
What can take the shapeless work, and what it can actually do
The connective layer has a property the hire never had. It is divisible. The chasing, the routing, the surfacing of the report that did not arrive, the first read on the exception before a human looks at it, this is work that can be picked up in fractions and handed off in fractions. A system can hold thirty percent of a coordinator and twenty percent of a reviewer at once without feeling that its job has shifted, because it has no job in the sense a person does. It has tasks.
This is the part worth being precise about, because the precision is where the honesty lives. What takes this work takes the operational substrate of the middle layer. It does not take the role. Accountability does not move, because a system has nothing at stake. Judgment about people does not move, because that was never operational work. When a hard exception arrives, a human still decides. What changes is that the human deciding is doing one coherent thing, not also chasing the four reports that had to arrive before the decision was possible.
The honest version of the benefit is narrow and worth stating plainly. The reports get reviewed, because surfacing and triage is exactly the kind of fractional work that can be handed off. The person who was holding the shapeless layer gets their actual role back, with edges. That is a real gain, and it is most of the gain.
What it does not fix, and why pretending otherwise is the trap
It is tempting to stop there and let the reader assume the rest follows. It does not. Two of the things that rot in a company this size do not get fixed by handing the connective work to a system, and a founder who believes they do will be worse off than before.
The first is SOP rot. A standard procedure goes stale because the business changed and no one updated the document. Handing that procedure to a system that can run it does not arrest the rot. It runs the stale version faster and more reliably, which can be worse than a human who at least senses that the steps no longer match reality. The document still has to be authored and kept true by someone who understands the work. That is definitional labor, and nothing automates it away.
The second is shadow processes. These are the workarounds that became permanent, the spreadsheet that quietly runs a function no one will admit is load-bearing, the approval that happens in a direct message and appears in no system of record. A system handed the official process cannot route around a thing it has no knowledge of. Shadow processes get resolved by being found and named, which means someone has to map how the work actually moves, not how it is supposed to move. That mapping is the work, and the automation comes after, if it comes at all.
So the line is clean enough to state. The fractional, connective, surfacing work can be made divisible, and doing that gives people back coherent roles and gets the reports reviewed. The definitional work of keeping procedures true and dragging the shadow processes into the light is a different kind of labor, and it has to happen first. A company that automates before it defines has simply hidden its structural problems behind a faster machine.
The smallest version of the whole point is this: the middle role was a bundle of fractional obligations pretending to be a job. Systems can help divide that work once it has been surfaced, named, and made legible. They cannot do that naming for you. Automating an undefined company does not create clarity. It just gives the confusion better throughput.