Picture a legal matter moving from a paralegal to an associate to a partner. It is the example Joshua Gans uses to open the August revision of a paper he first posted with the National Bureau of Economic Research in May.1 Each time the file changes hands, the person receiving it spends time getting oriented: reading the file, working out what was intended, and checking the work that came before. Gans prices that time at the receiver's wage, and it is one of the two costs that decide where job boundaries fall.1

What the handoff was holding together

The model is spare. Tasks are ranked by the expertise they need and grouped into jobs, and a job pays according to its most demanding task. Folding an easy task into an expensive job wastes expensive hours. Splitting the work across more people avoids that waste but makes every receiver pay the orientation cost. In the model, job boundaries settle where those two costs balance.1

Now cut the second cost. Gans looks at documentation standards, electronic records, and AI that briefs the next worker on a file, none of which performs a single task of the work itself.1 Jobs get redrawn anyway. With the number of jobs held fixed, the lower jobs widen and the top job narrows; as the orientation cost approaches zero, the number of jobs grows without limit and each one gets narrower.1 A tool whose only function is to brief the next person ends up changing who does what.

When a machine does the middle task and someone has to sign

Gans then works through a separate case. Take three tasks, low, middle and high in expertise. Before any machine arrives, a lower job handles the low and middle tasks and a top job handles the high one. A machine then takes over the middle task, and Gans assumes that checking its work must be done by whoever signs, "as when liability follows the signature."1 Under that rule the top job keeps its top task and its wage, but its list of duties grows and its average expertise falls. The lower job is left with only the low task, so the middle-expertise rung disappears, and in the model it disappears with or without the sign-off rule, across a range of conditions in which the machine is still worth adopting.1 Because the checking is done at the top wage, each unit of service costs more than it would without the rule, and output in the model is lower.1 Gans puts the finding in one line: "Bundle protection and role protection come apart."1

Two cautions belong here. The two results point in different directions, since cheaper briefing narrows the top job while automation with a sign-off rule lengthens it. And the three-task example is symbolic, with no numbers, in a working paper that has not been peer reviewed.1 I read both as mechanisms to look for, not forecasts.

A sign-off rule can keep the senior title intact while the job that used to sit beneath it quietly disappears.

What the top job may be carrying now

What follows is my reading, not a finding of the paper. An organization that watches titles would see little here. The partner is still a partner and, in the model, the partner's wage does not move; what moves is the list of tasks.1 The paper does not say what becomes of the person who held the middle role, and Gans leaves open whether that person becomes unemployed, takes a lower role or moves elsewhere.1 So the change would not show up in headcount by title or in the pay bands. It would show up in the week of the person at the top.

I covered a related pattern earlier this month, in an article on how saved writing time comes back as verification work. Gans models one structural version of it: when liability follows the signature, the checking lands on the top job.

A leader who wants to know whether this is underway does not need a new survey. In any function that adopted a tool that does part of the work and whose output a senior person must sign, ask that person what they now check that someone else used to do, and what they stopped doing to make room. Then ask a question the model does not address but I would ask anyway: where the next holder of that senior job is supposed to learn the middle work, if the rung that used to teach it is gone.

Which job in your organization got a longer list of duties in the last year without anyone changing its title?