Here is the number that stopped me cold: nearly nine in ten employees now use AI at work at least sometimes, but roughly 35% have gotten no AI training of any kind. We handed people a power tool and skipped the safety class. That, in one sentence, is the AI skills gap in 2026 — and it is a bigger threat to your results than any model, layoff headline, or competitor.
The data comes from Study.com’s State of AI Jobs and Skills 2026 report, and the picture it paints is one I recognize from the inside of real IT and operations teams. Adoption is basically universal. Structure is basically absent.
What the numbers actually say
Usage is everywhere, competence is not. According to the Study.com report, about 90% of employees use AI at least occasionally, yet 35% have received zero formal training. Of the people who did get training, only around 18% say it actually prepared them to work independently. Read that again: even among the trained, four out of five don’t feel ready to fly solo.
Mandates outrun standards. Roughly 21% of employees say AI use is a required part of their job and about 45% are encouraged or expected to use it. But only around 32% report a clear standard for what “good” AI use even looks like. So most people are being told to use it, and left to guess at what good looks like. Trial and error, no benchmarks, no feedback loop.
The payoff is real, which is the frustrating part. About 71% of workers already report saving time every week. The value is sitting right there. We are just leaving a big chunk of it on the table because we skipped the boring, unglamorous part: teaching people how to use the thing well.
The bottleneck in 2026 isn’t the AI. It’s the training gap around it. That’s the good news — because a training gap is something you can actually close.
Why this connects to the layoff headlines
You have seen the scary numbers. AI has been the single most-cited reason for job cuts for five straight months, and by mid-2026 it was named in well over 100,000 announced cuts — Challenger, Gray & Christmas put the cumulative 2026 tally north of 112,000 through July, roughly a quarter of all cuts. CNBC covered the same trend. It is real, and I am not going to wave it away.
But here is the nuance that matters. Research from Stanford’s Digital Economy Lab found that employment weakened in roles where AI mostly substitutes for human tasks, while it held flat or even grew where AI mostly augments people — especially for experienced workers. The damage is concentrated where AI replaces. The upside shows up where AI amplifies.
Put the two studies together and the strategy writes itself. The organizations getting hurt are the ones treating AI as a headcount-reduction lever. The ones pulling ahead are treating it as a capability multiplier for people who know how to drive it. And you cannot get the second outcome without training.
What this means for you
If you run a small business, the training gap is your opening, not your problem. You do not need an L&D department. You need to pick two or three tasks — drafting proposals, cleaning up customer emails, summarizing meetings — and write down what “good” looks like for each. That single page is more than what most companies have.
If you lead enterprise IT, you already know a mandate without a standard is how shadow AI and quiet data leaks happen. The 32% figure should be your target to move. A clear standard is a governance control, not just a productivity nicety.
Here is what I would actually do, in order:
Set a standard before you set a mandate. Write a one-page “what good AI use looks like here” — approved tools, what data never goes in, how to check the output. Boring. Essential.
Train for independence, not attendance. A lunch-and-learn that nobody can apply on Monday is theater. Tie every session to a real task people do every week.
Choose augmentation over replacement. Point AI at the drudgery inside a role so your people do more of the valuable part. That is where the Stanford data says the jobs — and the gains — actually hold.
Create a feedback loop. Share the prompts and results that worked. Peer learning closes gaps faster than any vendor course.
Don’t panic about the layoff numbers, and don’t ignore them either. The evidence is pretty clear: AI is not coming for the people who know how to use it. It is coming for the tasks nobody trained anyone to hand off well. The winners in 2026 will not be whoever bought the fanciest model. They will be whoever bothered to teach their people what good looks like.
Your next step this week: pick one task, write the one-page standard, and hand it to your team. That is the whole game. Close the gap before your competitor does.
News commentary by Brad Rowland — IT Infrastructure and Operations leader, automation builder, and AI implementer. Sources are linked inline.
