“A member asked me this week whether it was too late to start learning AI. I want to answer this properly. Not because the answer is complicated (it isn't) but because the coverage of this question is almost always either too alarming or too breezy, and neither actually helps.”
Is It Too Late to Start Learning AI?
No. It is not too late. We looked at where AI adoption actually sits across the population right now, not the tech industry, but the broader working population, and at what learning AI actually requires at this stage versus what people assume it requires. The honest picture is less alarming than the coverage suggests and more urgent than the reassurances suggest, and that combination is worth understanding properly. Most people have not started. The learning window is open. But "not too late" is not the same as "no reason to hurry," and conflating the two is where the cheerful nonsense comes in.
The question is not whether you are behind. The question is whether you are moving.
The numbers on AI adoption are worth knowing, because they tend to surprise people. As of early 2026, roughly a third of working adults use AI tools regularly, a third have tried them and stopped, and a third have not started at all. That last third is not a small, unusual group. It is a third of the population, which means the person asking "is it too late" is almost certainly not as far behind as they feel.
What has changed in the last eighteen months is the difficulty threshold. The tools have become genuinely easier to use. ChatGPT, Claude, and Gemini now require no technical knowledge to get real value from. You type what you need in plain English and get a useful response. Two years ago that was not reliably true. It is now. The people who started in 2023 had to work harder for worse results. The people starting now get a better tool on day one.
What has not changed is that experience compounds. Someone who has been using these tools for a year has built up an intuition for what works, what doesn't, and how to recover when it goes wrong. That knowledge comes from use, not from reading about it. Starting now means starting a year behind that person. Not an insurmountable gap. But a real one.
The honest editorial reaction: the window is open, the tools are ready, and the only thing that closes the gap is starting.
If you are in paid work of any kind (employed, self-employed, retired but still active in a field) the people around you who are using these tools are getting a practical advantage that compounds quietly over time. The gap between someone with a year of AI experience and someone starting today is not catastrophic, but it is real and it grows if you wait. The best moment to start was a year ago. The second best moment is using one of these tools for something you actually need done, today, before you close this tab.
If you are in paid work of any kind (employed, self-employed, retired but still active in a field) the people around you who are using these tools are getting a practical advantage that compounds quietly over time. The gap between someone with a year of AI experience and someone starting today is not catastrophic, but it is real and it grows if you wait. The best moment to start was a year ago. The second best moment is using one of these tools for something you actually need done, today, before you close this tab.
Your AI Prescription tells you exactly where to start based on your situation. Takes eight minutes. Kevin built it so you wouldn't have to figure this out alone.
Questions people ask
What should I actually learn first?
The most useful starting point is not a course or a tutorial. It is a task. Pick one thing you do regularly that feels repetitive, slow, or frustrating, and try using an AI tool to help with it. The learning that sticks comes from doing something real, not from watching someone else do it.
How long does it take to get genuinely useful at this?
Most people get real, practical value from AI tools within the first two or three sessions, not because they have mastered anything, but because the tools are accessible enough that the learning curve is shallow at the start. The deeper skill, knowing which tool to use, how to write prompts (the instructions you type into the tool) that get reliable results, and how to spot when the output needs work, develops over weeks and months of regular use.
I tried ChatGPT once and it wasn't useful. Does that count as starting?
Trying it once with a vague prompt and getting a vague result is how almost everyone starts, and it is a genuinely misleading introduction to what the tools can do. The difference between a poor result and a useful one is usually in how specifically you describe what you need. One more try, with a real task and a specific ask, is worth more than reading ten articles about AI.