How to Learn AI at Work When Nobody's Training You
There’s a quiet gap in the workplace right now that nobody likes to talk about. Around 77% of employers say they plan to reskill their workforce for AI. But only about 13% of employees report having actually received any AI training. The intention is nearly universal; the delivery is almost nonexistent.
If you wait for your company’s official AI training program to show up, you may be waiting a long time — and quietly falling behind peers who didn’t wait. The good news: real Not knowing how the technology works under the hood, but knowing how to use it well — what it’s good and bad at, how to get useful output, when to trust it, and when to double-check. It’s a practical skill, closer to “knowing how to search effectively” than to computer science, and you build it by doing, not by studying. is something you can build yourself, for free, in the course of your normal work. Here’s a path that actually works.
Step 1: Use it daily on real work — not toy prompts
The single biggest driver of fluency is reps on real tasks. Playing with AI by asking it to write limericks teaches you almost nothing useful. Using it on the actual email, the actual report, the actual analysis you owe someone by Friday teaches you fast, because you can instantly tell whether the output is good enough.
Pick one recurring task you already do and commit to running it through AI for two weeks, even when it’s slower at first. That friction is the learning.
Step 2: Learn the two mental models that explain everything
You don’t need to understand the math. You need two concepts, because almost every AI quirk traces back to them:
- Context is everything. The model only knows what’s in front of it in the conversation — it has no memory of your business unless you paste it in. Most “bad AI output” is really an information problem you can fix.
- It can be confidently wrong. A When a model states something false with complete confidence, because it’s built to produce a plausible-sounding answer rather than to look facts up. Knowing this reflex exists is what turns you from someone who trusts AI blindly into someone who knows exactly which parts to double-check. isn’t a rare glitch; it’s a predictable behavior you learn to anticipate.
These two ideas are unpacked in plain English in Giving AI Context: The Skill That Actually Makes It Useful. Internalize them and most of AI stops being mysterious.
Step 3: Build one reusable thing
There’s a leap in understanding that only happens when you stop typing one-off prompts and build something you reuse. Turn a task you repeat into a saved recipe — a report, a summary, a checklist. Watching yourself go from “re-explaining it every time” to “run it and review” teaches you what these tools are actually for.
Two concrete walkthroughs to copy: How to Write a Recurring Report With AI in 10 Minutes and, if you’re a bit more technical, Build Your First Command — Then Grow It Into a Skill.
Step 4: Develop judgment about when to trust it
The most advanced part of AI literacy isn’t getting better output — it’s knowing when not to rely on it. That means building an instinct for which tasks are safe (easy to check, low stakes) and which aren’t (hard to verify, high consequence). That judgment is the difference between AI saving you time and AI quietly creating expensive mistakes.
The frameworks for this are in How to Pick Your First AI Task at Work and Why AI Doesn’t Always Save Time.
Step 5: Keep a wins-and-fails log — and share it
Spend thirty seconds after each attempt noting what worked and what flopped. Two things happen. First, you build a personal playbook faster than any course would give you. Second, when your company’s training does finally materialize — or when a colleague is struggling — you’re the person who already knows, which is a quietly valuable place to be.
The takeaway
The training gap is real, but it’s also an opportunity. While most people wait for a program that may never come, you can build genuine AI literacy the way it’s actually built: by using the tools on real work, understanding the two mental models behind their behavior, making one reusable thing, and developing the judgment to know when to trust the output. Nobody’s coming to train you. That’s exactly why doing it yourself is such an edge.
Sources: The ~77% of employers planning to reskill or upskill their workforce for AI is from the World Economic Forum’s Future of Jobs Report 2025; the finding that only around 13% of workers have actually received AI training comes from Randstad’s workforce research.
Related: Giving AI Context: The Skill That Actually Makes It Useful for the mental models, and AI Anxiety at Work: You’re Stressed and More Productive. Both Are True. for why building this fluency is the best answer to the worry.