The AI Skills Gap Isn't a Training Problem.

The overview of our 2027 HR priorities flagged the AI skills gap as the third thing SMBs can’t afford to ignore next year - and the one most likely to get misdiagnosed. Most of what’s written about this topic assumes a company big enough to have a learning and development function that can scope a curriculum, hire a vendor, and roll out formal training. If that’s not your company, this post is for you.

The gap is real, and it’s not evenly distributed

Start with how uneven AI adoption already is inside a typical organization. Research on workplace AI readiness has found that only 24% of individual contributors strongly agree their employer actually prepared them to use AI effectively - while 77% of managers believe their teams are ready. That’s not a small measurement error. That’s two groups of people, in the same company, describing two different realities, and the group with less visibility into day-to-day work is the one that thinks everything’s fine.

The training gap behind that perception gap is stark: only about 16% of individual contributors and 23% of managers say they received any preparation before a new AI system was rolled out to them. Tools are shipping ahead of any instruction in how to use them, which means whatever skill exists on a team right now developed in spite of the rollout process, not because of it.

And where guidance does exist, it’s inconsistent by design rather than by policy: roughly 30% of employees report that AI guidance varies team to team with nothing written down, and 21% say they’ve gotten no guidance at all. There’s no single villain in that data - no policy anyone consciously chose. It’s just what happens when nobody owns the question.

What this actually looks like inside a 100-person company

Play this forward at SMB scale and you get a very specific, very familiar picture: one or two people on the team have quietly become the de facto AI expert - not through any formal designation, just because they were curious early and kept experimenting. Everyone else is operating somewhere between cautious and avoidant, not because they don’t want to use the tools, but because nobody has told them whether it’s expected, whether it’s allowed, or whose job it is to teach them.

That uncertainty isn’t neutral. It’s actively slowing adoption, and it’s producing a specific kind of quiet anxiety: research on this exact dynamic found that roughly a third of individual contributors and about a quarter of managers believe AI will outright replace some jobs, including their own. People don’t raise their hand and ask “am I supposed to be learning this?” when they’re already worried the honest answer might cost them their job. They just quietly fall further behind, and the gap compounds.

Why this is a retention problem, not a productivity problem

Here’s the reframe that matters: the informal AI expert on your team is accumulating real leverage and visibility that nobody has formally recognized or compensated. They know it, even if they haven’t said it out loud. Meanwhile, everyone else is falling behind on a skill that’s rapidly becoming table stakes, watching that gap widen in real time, with no clear path to closing it and no one whose job it is to help them close it.

That combination - unrecognized expertise on one side, quiet anxiety about obsolescence on the other - is exactly the kind of gap that turns into a resignation. Not dramatically, and not with an exit interview that names AI directly. It shows up as your best informal expert leaving for a place that will actually credit what they know, or as a solid mid-tier performer who quietly starts looking elsewhere because they can feel themselves falling behind and nobody offered them a way to catch up.

There’s a manager-capacity angle here too, and it connects directly to the other two priorities on our list of 3 Priorities SMBs Can’t Ignore in 2027. Gallup’s research shows that when a direct manager actively champions AI use, employees are dramatically more likely to say the technology has improved their work - multiple times more likely, by some measures, to report real productivity gains. Manager engagement isn’t just about one-on-ones and goal setting. It’s turning out to be the single strongest predictor of whether a team actually adopts the tools it’s been given. A stretched-thin manager with no bandwidth to engage with AI is a team that stays stuck at the informal-expert-and-everyone-else stage indefinitely.

What actually helps — and what doesn’t

The instinct to respond to a skills gap with a training rollout is understandable, and for most SMBs it’s the wrong move. You don’t have the L&D function to build a curriculum, you don’t have the budget to buy one off the shelf and sustain it, and a mandatory training module nobody asked for tends to produce compliance, not skill.

What actually helps is naming the gap out loud before it hardens into resentment:

That last point is the one that matters most, and it’s not a technology purchase. It’s a manager conversation - the kind of thing that surfaces naturally in a one-on-one that already has room for “how’s work going” rather than one narrowly scripted around status updates and deadlines. If your one-on-one structure only ever asks about deliverables, this conversation has nowhere to happen. If it leaves room for the actual texture of someone’s week, it comes up on its own.

Where to start

If you already sense an uneven AI adoption gap on your team - and if you’re honest, you probably do - start by naming it, not by announcing a training initiative. Ask your managers to raise it directly in their next round of one-on-ones: not “are you using AI,” which invites a defensive yes, but “what have you tried, what’s worked, what hasn’t” - a question that makes it safe to admit “not much yet” without that admission feeling like a confession. Do that consistently for a quarter, note who comes up as the informal experts, and you’ll have a real map of where the gap actually is instead of a guess. That map is worth more than any curriculum you could buy before you have it.