Freeze backfills to fund your AI budget.
When someone hands in their notice, the reflex is to dust off the old job description and post it again by the end of the week. The work was there yesterday, so the seat should be filled tomorrow. Before that post goes up, there is a better question to ask, and it costs nothing but a few weeks: is this still the job?
A departure is a design decision.
Most job descriptions in a typical organization were written before anyone on the team used AI for real work. Since then, parts of many roles have changed shape. Drafting a first version of a letter, summarizing a long thread, pulling notes into a report, entering data from a form, doing the first pass of research on a vendor: these now take a fraction of the time they once did for someone who uses the tools well. The description in the HR folder still lists them as if each one fills a morning.
A resignation is the one moment when you can redesign a role without changing anyone's current job. Nobody is being replaced. Nobody is being told to absorb more work on the spot. You are deciding what the team needs next, with an open seat and a budget line that has already been approved.
In most organizations, getting new money for AI training or tools means a request, a justification and a wait. The salary for an open role is money the organization has already agreed to spend.
Pause before you post.
A backfill freeze holds the seat open for a set period, usually a month or a quarter, while you watch what happens to the work. You may well hire at the end of it. Pick the length up front and put the review date on the calendar, so the freeze has an end and nobody mistakes it for a quiet cut.
During that window, the work sorts itself. Some of it turns out to matter a great deal, and you feel its absence within days: invoices sit, the intake queue grows, a client asks why nobody called back. Some of it gets picked up by tools the team already pays for, once someone sits down and tries. And some of it turns out to have been habit, a weekly report nobody read or a spreadsheet kept in parallel with the system that replaced it.
The salary for an open seat is the one AI budget most organizations have already approved.
Ask the manager to keep a short log of what goes undone, what gets done another way and who is carrying the extra load. Without that log, a freeze turns into a few tired colleagues covering quietly until someone burns out, and you learn nothing about the role.
Three ways to fill the seat.
At the end of the window, the log points to one of three outcomes.
You can refill the role as it was. The work is real, it is mostly human, and it was not keeping up even before the person left. A town office clerk who spends the day at the counter with residents is a fair example. Hire, and don't apologize for it.
You can refill it as a new role. The work is still needed, but the shape is different now. Perhaps a coordinator at a clinic used to spend half the week building schedules and chasing confirmations, and the team has since found that most of that can be drafted and sent with a person checking it. The new hire still coordinates, but the description should say what they will actually do, which tools they will work alongside and what they are responsible for checking.
Or you can hold it. The work is being handled well enough by the people and tools already in place, and the extra load is fair. Keep the seat open, set another review date and move some of the money to where it does more good. If nobody can say when the held seat will be looked at again, it has stopped being a decision.
Put the savings back into the team.
The money from a held seat, or from the difference between an old role and a redesigned one, is the most natural AI budget a smaller organization has. It needs no new line item. Spend it on the training, tools and protected time the remaining team needs to keep up with the work they absorbed.
This part is not optional, and it is where freezes most often go wrong. If the people who made a seat unnecessary see none of the benefit, they draw the obvious conclusion: getting good with AI means more work for the same pay, and possibly fewer colleagues. They will stop finding savings, and they will be right to. Agree with the team, openly, how the money from a held role gets spent, and show them where it went at the next review.
In practice that might mean paid hours for a nonprofit's program staff to learn the tools on their own reports, a raise for the person who now covers two functions, or an outside session for a manufacturer's front office on the workflows that changed. The specific choice matters less than the team seeing that the savings came back to them.
A four-step backfill review.
Run this for every open role, and for every departure you can see coming in the next two quarters.
- Ask the manager where the work goes if the seat stays empty for a quarter, and have them write it down task by task.
- Mark which of those tasks AI tools the team already has could take on today, which they could help with under a person's review, and which they can't touch.
- Decide: refill as it was, refill as a new role, or hold for a set period with a review date.
- Add up the held or saved salary and agree with the team, in writing, how that money gets reinvested in training, tools, time or pay.
Keep the notes from each review together. After a year you will have a record of how your roles have changed, which is a better guide to the next hire than any job description in the folder.
