Redesign one role, not the org chart.

When leaders start thinking about what AI means for how their organization is staffed, the conversation tends to jump straight to the org chart. Which departments shrink, which ones merge, what the structure looks like in three years. It is an understandable place to start and almost always the wrong one, because nobody can redesign a structure they can't see the work inside of.

Why the org chart is the wrong unit.

An org chart shows who reports to whom. It says nothing about how a Tuesday morning actually goes for the person in each box. Two people with the same title can spend their weeks very differently, and the boxes hide the part that AI changes, which is the individual tasks.

Large redesigns also stall for a simpler reason. They ask a lot of people to agree on a future none of them can picture, and they make everyone nervous at once. A plan that moves twenty boxes gets debated for months. A plan that changes how one coordinator handles her mornings can be tried by Friday.

So shrink the unit. Pick one role, describe it honestly, and rewrite it with AI in it. If the rewrite works, you have a pattern you can repeat on the next role, and the structure starts to change from the bottom up, on evidence.

Choose the role that sits closest to your goal.

Start from the outcome the organization cares most about this year, stated in plain terms. Shorter turnaround on client proposals. Faster intake at a clinic. Fewer late reports to a funder. Then ask which role does the most work that feeds that outcome, and pick that one.

The temptation is to pick the role with the most obvious busywork. That can work, but a role tied to a real goal gives the exercise a reason to exist beyond tidiness, and it makes the result easy to judge later. In a school district, that might be the enrollment coordinator. At a small engineering firm, it might be the project manager who assembles proposals.

A plan that moves twenty boxes gets debated for months; a plan that changes one person's mornings can be tried by Friday.

Write the role as an ordinary week.

The written job description is the least useful document here. It was written for hiring, it lists responsibilities in the abstract, and it is usually out of date. What you want is a narrative of the actual week, written with the person who holds the role.

Say the role belongs to an operations coordinator named Dana (a made-up example). Her mornings go to the shared inbox: sorting vendor emails, answering the same five questions from staff, and copying order details into a spreadsheet. Early afternoons go to chasing missing paperwork. Late in the week she builds a status report by pulling numbers from three systems by hand. Somewhere in there is the work she was actually hired for, like negotiating better vendor terms, and it gets whatever time is left.

Write it in that register: what fills each part of the week, what is repetitive, where things wait on other people, and what keeps getting pushed. That last item matters most. The work a person never has time for is where most of the value in a redesign ends up.

Rewrite the same week with AI in it.

Now write the week again. Go task by task and ask what AI could draft, sort, summarize or check, and what the person would still need to decide. In Dana's rewritten week, the inbox arrives pre-sorted with suggested replies to the common questions, which she reviews and sends. Order details are pulled into the spreadsheet for her to confirm. The status report is assembled from the three systems and she edits the summary instead of building it.

Then write down what she does with the hours that come back. If that line stays blank, the redesign has only removed work, and people notice. The rewrite should name the higher-value work by name (the vendor negotiations, the onboarding guide nobody has written), so the person can see a better version of their own job in it.

Do this with the person in the room. They know which shortcuts are safe and which ones would cause trouble, and a role someone helped redesign is a role they will actually try to work in.

List what gets in the way, then rank the changes.

As the rewrite takes shape, keep a separate list of the things that block it. Old systems that won't export data. A spreadsheet nobody trusts. A tool the organization already pays for but nobody was trained on. A policy question about what client information can go into an AI tool. Some of these turn out to be larger projects in their own right, and it is better to see them now than halfway through a build.

Then rank the changes. Start with the ones that are quick and low-risk even if they are small, because early wins give the person and their manager a reason to keep going. Leave the changes that depend on cleaning up data or replacing a system for later, and say so plainly so nobody expects them next week.

A one-role redesign worksheet.

Fill this in with the person who holds the role, in one sitting of about two hours.

1. Goal: the one organizational outcome this role most affects, in a sentence.
2. Role: title, and the name of the person filling in the sheet with you.
3. The current week: a short narrative of how each part of the week is spent.
4. Waste: tasks that are repetitive, re-keyed by hand or waiting on someone else.
5. Missing work: what this person would do with more time, in their words.
6. The rewritten week: the same narrative with AI handling or drafting specific tasks, and what the person still decides.
7. Reclaimed time: where the freed hours go, named task by task.
8. Blockers: systems, data, tools, training or policy questions in the way.
9. First three changes: the quickest, lowest-risk items to try in the next two weeks.
10. Check-in date: when you will sit down again to see what changed and pick the next three.