Will AI replace your staff?

Our answer for companies of 20 to 500 people. What AI takes over inside each job, what stays with people, and how to tell your team about it.

For a company of 20 to 500 people, we think the answer is no. AI takes over the repetitive admin inside each job, and the same team gets more done. Some tasks will disappear and a few roles will change a lot, so tell your staff plainly which ones.

What owners tell us

The news stories are about layoffs. The owners we talk to describe a different problem. One of them said it to Taylor like this.

"We're a small firm and we do amazing work for amazing clients and they want more of us. The only way they can get more of us is if we're not behind a computer."

We hear versions of that often. The team is good, clients want more of it, and too much of the week goes to moving information from one place to another. Taylor's view is that a small business needs the same people producing three times as much. Julian thinks that's achievable.

Taylor wrote in April that AI isn't here to replace your people, and he put the word "yet" in brackets after it. We don't know what the tools will do in five years, so this article sticks to what we see them doing now.

What AI takes over

On a first call with a skeptical owner, we asked her to bring a task her team fights with every week. She brought a messy file, spread across several locations, that took five hours a week to wrestle into shape. We put it through Claude on the call and it was done in five minutes.

A printing and manufacturing company had a vendor spreadsheet that someone reformatted for eight hours every week. One session brought that down to 10 minutes. A project manager at a real estate advisory firm told us research for each project file went from four hours to 45 minutes.

How we decide what stays human

We interview every person on the team, one on one, and sort each piece of their work into three groups.

GroupWhat it meansExamples from our work
Stays humanJudgment, relationships, and final decisionsDeciding what a report means for the client, approving anything that goes out
AI assistsAI does about 70 percent and a person finishesResearch and assembly for a report, a drafted email the sender reads and approves
AI does itRepetitive work with clear rules, checked by exceptionReformatting a vendor spreadsheet, copying data between systems

The sorting becomes a roadmap, ordered by priority. Staff see their own time sinks on it, because the list came from their interviews.

Which jobs change

Tasks do go away. If someone spends most of their week on work in the third group, their job will look very different in a year.

Say so early. We like to open with a 30-minute kickoff for the whole team that covers what the first month will look like and what people should expect from us. The survey we send before interviews asks each person what makes them nervous about AI.

Then tell people where the time is going. At a flagship hotel we trained, sales reps put their day at 70 percent admin and 30 percent selling. A rep who hears that the goal is more of the week spent selling has a reason to help.

Hiring and offshore teams

Clients ask us whether we would tell them not to hire. We would, and we have said so on calls. If AI can cover a role for now, the hire can wait until the work is clearly there.

One owner came to us wanting to scale back an overseas admin team over time, and was weighing a product that promised a team of agents to do it. We think the goal is reasonable. Julian's advice was to do it in stages and build trust first, with a person reviewing the output. He also said he didn't think agents were the place to begin, and that an automation which flags the work and leaves the decision to a person should come first.

Locking AI down pushes people out

The staffing risk we'd worry about more runs the other way. Companies restrict AI to one approved tool, or ban it, and their most motivated people notice. Employees at those companies have told us they resent it. Some have started looking for employers who will train them on the newest tools.

We think of AI access as a retention issue. Staff understand the security concern. They also know that AI skills are what will keep them employable, and they will go where they can build them.

What to measure

Measure output for the same team. Count the jobs quoted, reports shipped, or clients served, along with the hours each person gets back in a week.

Skills are worth measuring too. In one 10-week program at a Canadian restaurant group, the AI champions from each department rated their own skills at 2.5 out of 5 at the start and 4.5 at the end. Those people went back to their departments and taught everyone else.

When staff don't use what was built, we treat that as our failure. The fix is usually to build around the way people already work, so they get the benefit before they're asked to change anything.

Where to start

Ask each person on your team which weekly task they would hand over first. You'll get a list of the admin nobody wants, in their words, and a first sense of how much time is sitting in it. Share the list with the team before you automate anything on it.

If you want help turning that list into a roadmap, book a free audit and we'll walk through it with you.

More guides.

Find out what we'd build for you.

The free audit is a 30-minute call. We ask how your team works today and tell you where we'd start.

Book a free audit ↗︎