Demand for AI skills grew 7x in two years. Employers still can’t explain what they’re looking for.
I’ve worked with salespeople and marketers who considered themselves AI-savvy. They used ChatGPT daily. They could talk about prompts. On paper, they’d check the “AI fluency” box on any job application.
Then I looked at what they were actually doing.
The salesperson was generating outreach without any brand voice or messaging context loaded. The output was close enough to sound professional — but it was a slippery slope toward off-brand promises and generic pitches that could have come from any company. The marketers were churning out blog posts for volume, not authority — no EEAT strategy, no specialization, just AI-generated content that wouldn’t build trust with readers or search engines.
They weren’t doing anything wrong, exactly. They just didn’t have a system. No centralized context. No framework for consistent prompting. No way to track what was working. They were using the tool without the infrastructure that makes the tool useful.
I use Claude Projects extensively for exactly this reason — brand voice, messaging frameworks, SOWs, and past work all loaded in one place. When I prompt, the AI already knows the context. The output is consistent because the inputs are consistent. That’s the difference between “using AI” and having a system for AI.
This is the invisible gap in “AI fluency” — and it’s everywhere.
The Fastest-Growing Skill With No Definition
McKinsey’s latest research puts a number on the shift: demand for AI fluency — the ability to use and manage AI tools — has grown sevenfold in two years, making it the fastest-growing skill in U.S. job postings.
Seven times. In two years.
Yet when you look at what employers actually want, the picture gets murky fast. Some companies mean “curiosity and willingness to learn.” Others want full technical implementation. A health-tech firm told applicants they’d be evaluated on how creatively they used AI to identify growth opportunities. Zapier has gone further, creating an internal framework that ranks employees from “Unacceptable” (resistant to AI tools) to “Transformative” (reimagining processes and building new value with AI).
The gap between “Capable” and “Transformative” in Zapier’s model? A capable employee uses AI to help draft a report. A transformative one builds a tool to automate the report entirely.
That’s not a small difference. And most job postings don’t specify which they want.
The real question isn’t “do you use AI?” It’s “do you have a system for using AI?” — and almost nobody is asking that in interviews.
The Economics Are Real — Even If the Definitions Aren’t
Here’s what we do know: AI fluency pays.
PwC’s 2025 Global AI Jobs Barometer — based on analysis of nearly one billion job postings across 24 countries — found that roles requiring AI skills now command a 56% wage premium over similar roles without AI requirements. That’s more than double last year’s 25% premium.
But here’s the twist: the jobs most exposed to AI are seeing the slowest growth in new postings.
Between 2019 and 2024, U.S. job postings for highly AI-exposed roles (software developers, finance managers) grew just 1%. Meanwhile, roles least exposed to AI — bricklayers, food preparers — saw 20% growth.
The explanation is simple: AI-fluent workers are so productive that companies don’t need as many of them. PwC found productivity growth nearly quadrupled in AI-exposed industries since 2022. When each employee generates more output, you hire fewer employees.
This isn’t job destruction. It’s job consolidation. And for SMBs, it changes the hiring calculus entirely.
What This Means for SMBs
If you’re a small business owner trying to hire, you’re facing a strange situation: everyone says they want “AI skills,” but the candidates who have them are expensive, and you’re not even sure what you’re looking for.
Here’s the opportunity buried in that confusion.
You get to define it. Unlike enterprises with HR committees and standardized job descriptions, you can decide what AI fluency means for your business. You don’t have to match some unstated industry standard that even recruiters can’t articulate.
Start by asking: What are the three to five tasks where AI could actually help this role? Then evaluate candidates on whether they have a system for using AI on similar work — not whether they’ve heard of the right tools.
Zapier’s framework is actually useful. Their four-tier model (Unacceptable → Capable → Adoptive → Transformative) gives you a concrete way to think about AI skills without requiring a computer science degree:
- Capable: Uses popular AI tools, has some hands-on experience
- Adoptive: Embeds AI into personal workflows — prompting, chaining tasks, automating repetitive work
- Transformative: Rethinks processes entirely, creates new value
For most SMB roles, you probably need “Adoptive.” You’re not hiring someone to build custom AI infrastructure. You’re hiring someone who won’t waste three hours on a task that AI could do in ten minutes.
The premium is real, but so is the leverage. That 56% wage premium exists because demand outstrips supply. But here’s what the data also shows: AI-fluent workers are more productive, which means you might be able to afford one excellent hire instead of two mediocre ones. The math changes when output per person goes up.
The Equity Problem Nobody’s Talking About
There’s a harder truth here. Not everyone has equal access to AI tools, training time, or the kind of job that lets you experiment. If employers aren’t careful, “AI fluency” becomes another vague gatekeeping term that screens out qualified candidates who simply haven’t had the opportunity to learn.
For SMB owners, this is both a risk and an opportunity. You can build AI capability internally rather than demanding it upfront. Hire for learning ability, then invest in training. The 56% wage premium partly reflects scarcity — and scarcity you can solve internally costs less than scarcity you try to hire around.
The Bottom Line
AI fluency is real. The wage premium is real. The productivity gains are real.
But the definition? Still a moving target.
For SMBs, that ambiguity is actually an advantage. You’re not bound by enterprise HR processes that can’t adapt fast enough. You can define what AI skills matter for your specific workflows, evaluate candidates on actual use cases rather than buzzwords, and build capability internally instead of paying the premium to hire it.
The companies that figure this out first won’t just hire better. They’ll pay less for more productivity — while their competitors are still arguing about what “AI fluency” even means.
Not Sure What AI Fluency Should Look Like for Your Team?
This is exactly what I help SMBs figure out. At Pallas Advisory, I work with small businesses and marketing teams to build the systems that turn “we use AI” into actual results — centralized context, prompting frameworks, and workflows that produce consistent output instead of generic slop.
If you’re hiring and don’t know how to evaluate AI skills, or you’ve got a team that’s using AI without a system, let’s talk.
Sources
- McKinsey Global Institute, “Agents, robots, and us: Skill partnerships in the age of AI” (November 2025)
- PwC, “2025 Global AI Jobs Barometer” (June 2025)
- PwC Press Release on AI Jobs Barometer findings
- Zapier AI Fluency Framework via Fisher Phillips
- ERE Media on AI fluency wage premium and hiring slowdown
