There's a version of "using AI" that means opening a chatbot to draft an email faster. That's real, and it's useful, but it's also the smallest possible version of what's available. I learned the difference between using AI and building with AI the way I learn most things — by needing it to be true, not by reading that it was.
For years I was running The JetLyfe — sourcing, marketing, customer service, content — almost entirely by myself, on top of a full-time job. There was no team to delegate to. If something didn't get automated, it didn't get done, because there weren't enough hours in a week to do everything a small business actually requires by hand. That constraint is what pushed me past "AI as a convenience" into something closer to "AI as infrastructure."
The distinction that actually matters
A tool decision asks: what's the fastest way to get this one task done? A systems decision asks: what should never require a human again, and what's the structure that makes that true?
Those produce completely different outputs. If you're making a tool decision, you ask AI to write a product description once. If you're making a systems decision, you build a repeatable process — a prompt structure, a set of brand guidelines fed into every request, a review step — so that every product description meets a bar without you personally writing or even reading most of them. The first saves you twenty minutes. The second changes what your business can scale to without adding headcount.
The question isn't "can AI do this task." It's "should this task exist as a task at all, or should it exist as a system that runs without me."
Where this shows up in the work
At IKEA, the same logic applies to coaching. A tool decision would be using AI to write one role-play script for one rep before one call. A systems decision is what I actually built: a library of scenario structures, reinforcement scripts, and coaching templates that scale across a team of twelve and keep producing new, realistic practice material without me hand-writing every scenario from scratch. The output isn't "AI helped me today." It's "the team has a coaching system that didn't exist before, and it keeps running."
The same thing happened with grant writing for The JetLyfe. Writing one strong grant application with AI assistance is a tool decision. Building a reusable structure — the questions every funder tends to ask, the proof points that consistently work, a process for adapting that core material quickly to a new opportunity — turned a task I used to dread into a system I could execute against whenever a new opportunity appeared, instead of starting from zero every time.
The leadership implication
This is also, I'd argue, an executive skill more than a technical one. The people getting the most out of AI right now aren't necessarily the most technical people in the room — they're the people who are good at identifying where a repeatable structure is hiding inside a messy, manual process. That's the same skill that makes someone good at building an org chart, a training curriculum, or a sales process. AI didn't introduce that skill. It just made it dramatically more valuable, because the cost of actually building the system dropped close to zero.
The practical takeaway, if you're a leader trying to figure out where AI fits in your organization: stop asking "where can we use AI" and start asking "which of our repeated, manual tasks deserve to become a system instead of a habit." The tasks that answer "yes" are where AI earns its place. The rest is just a faster version of typing.