A yes that changes nothing
When an owner asks the crew or the office whether they use AI, the reply describes habits, not the business. Someone drafts emails with it, someone else summarizes calls, and next month the schedule, the estimates and the payroll look exactly the same. You learned who is curious. You did not learn where the work could be organized differently.
Shopify tried a formal version of that question. In an April 2025 memo, CEO Tobi Lütke wrote: "We will add AI usage questions to our performance and peer review questionnaire." The peer question asked how readily a colleague arrives with an AI prototype. In a Sources podcast episode published October 28, he told Alex Heath the answers "didn't matter."
The rule that kept working
The same memo carried a second rule. Before requesting more headcount, teams had to show why AI could not do the work. Lütke said that rule and the AI expectations now happen on their own. "I don't have to enforce these kind of things," he said, and added that he does not know whether the AI question is still in the current review.
The difference is where each rule sits. A review question arrives after the work is done and measures attitude. A headcount rule arrives before money is committed and asks for an answer about the job itself. Lütke saw the review question as a way to set expectations that can "give a company a reputation to live up to." Today, he said, those expectations happen organically, as part of how the company works.
Hiring and tools, decided together
In a roofing, renovations or home services company, the usual sequence is to feel the overload, post the job, and only later wonder whether a tool could have covered part of it. That order shapes the hire around tasks nobody questioned, and the tools end up bolted onto a role that was already defined.
Writing down what the new person will do that AI cannot reverses the order. The human part tends to be easy to name: walking a roof with a homeowner, handling a tense callback, reading a job site, earning trust on the phone. Whatever is left on the list becomes something to test with a tool first, in the same conversation, before it turns into a job description.
What changes this quarter
Before the next opening is posted, whoever approves hiring adds one written paragraph to the request: what this person will do that AI cannot, and what the team will try with AI first. The office manager drafts it, the owner reads it in the weekly meeting, and the hire and the tool budget are approved together, not one after the other.
Within the first week the conversation gets sharper. Job descriptions name the human work plainly, the team tests tools against real tasks instead of debating them, and the quarterly plan shows people and software as one set of decisions. Nobody needs to be asked whether they use AI, because the question is already built into how the company grows.