There is a common saying in investing: it is not just about how much capital you have, but how much time you have.

The best investors do not win simply because they deploy the most money. They win because they give their decisions time to compound. Small gains, repeated over long enough periods, turn into something much larger than they first appear.

I think the same is true for AI.

Right now, we’re still so early in the AI adoption era: a lot of people talk about AI adoption as if it requires a full organizational reinvention, complete with new infrastructure and an entirely new operating model. You see early adopters spinning up local models, wiring together agent stacks, and building elaborate internal setups that look more like research labs than work tools.

PS: if you ever feel behind in your AI use and skills, I look at this image often to remind myself that we’re really ahead of the curve here.

For the vast majority of teams, the real advantage in AI is not complexity. It is time.

We have seen this pattern before. When electricity first entered factories in the late nineteenth and early twentieth centuries, many manufacturers treated it as a one-for-one substitute for steam power. They replaced one central engine with one central motor and left the rest of the system intact. The technology changed, but the workflow did not. The real gains came later, when factories were reorganized around what electricity actually made possible: distributed power, more flexible layouts, faster production, and new forms of coordination.

In other words, the breakthrough did not come from installing the new tool. It came from learning how to work differently because of it.

Something similar happened with spreadsheets. Companies did not become spreadsheet-native overnight, and no one disappeared for a week and returned with an entirely rebuilt finance organization. Instead, adoption happened in smaller, more ordinary ways. One person used a spreadsheet to model a forecast. Another used it to clean up a reporting process. Another stopped doing a repetitive calculation by hand. Over time, the people who kept using the tool developed intuition for what it was good at, where it broke, and how it changed the speed and shape of work.

The advantage was not that they understood everything immediately. The advantage was that they stayed with it long enough for fluency to build.

That is where I think we are with AI now.

Last week, I spoke with a friend who is deep in the AI world, the kind of person building systems, infrastructure, and the whole stack. He showed me workflows that were powerful, sophisticated, and honestly a little overbuilt for what I needed. I have run into this dynamic before with consultants: what feels obvious to the builder is not always aligned with the actual scope of the problem. The person closest to the tool often sees its maximum potential. The person doing the work usually just needs a better way to get through Tuesday.

My goal was not to build an AI-native organization overnight. My goal was to make my team more effective now.

And the fastest path there was not a full system overhaul. It was a series of small, targeted tools for specific workflows: one prompt that saves ten minutes, one workflow that removes a repetitive task, one experiment that shows you where the tool is actually useful, and then another, and another.

That is the part people miss when they talk about "adopting AI." You do not need to adopt AI all at once. You need to start using it.

Because every small use creates two returns at the same time. First, it produces an immediate gain. Maybe you save time. Maybe you get unstuck faster. Maybe you move through a draft, a report, or a planning document with less friction. But second, and more importantly, it builds judgment. You learn what kinds of prompts work, which tasks are worth delegating, which ones still need your brain, where the tool sounds smart without actually being right, and how to shape its output instead of just reacting to it.

That learning compounds.

The people who benefit most from AI will not necessarily be the ones with the most technical setup. They will be the ones who have spent the most time in contact with the tools, building taste, judgment, and instinct through repeated use.

This is the prompt I use and paste into the LLM I use most often to get me out of a rut:

Because the real moat in AI is not early mastery. It is sustained participation.

The winners will not be the people who understood everything on day one. They will be the people who kept showing up, kept trying things, kept noticing what worked, and kept letting those small wins build on each other. Open the tool, try one real task, break it, fix it, and try again. Let the gains be small at first. That is how compounding works.

One prompt, one workflow, one useful idea can, over time, turn into something much bigger than you expected.