Are you an AI Illiterate?
Are You an AI Illiterate?
Jul 21, 2025The Quiet Battle Over AI Literacy—and the Fortune It Will Mint for the Few.
The Brief
Early this year, when the Seattle sun was still hidden behind a gray, I tasked three designers with an identical task to imagine a new product. “█████ ████ ██████ ██ ███ ███████.” The only constraint was each had two weeks until they can tell me if we should invest and if the concept has legs. Each worked on the problem independently, and since they worked across different teams they weren’t aware of each others pursuits.
All three designers have similar design school backgrounds, are extremely talented, with no prior coding experience. The only difference between them was the total years of experience.
Two weeks later, it was time to review their outputs, which were in drastically different formats.
| Designer A | Designer B | Designer C | |
|---|---|---|---|
| Years of Experience | 17 | 13 | 4 |
| Deliverable | ~10 core concept screens in Figma; concept narrative and MVP definition | Five‑page strategic memo; TAM analysis; moat diagram; MVP feature breakdown | Live web demo; link sent two days prior to deadline |
| AI Usage | Used ChatGPT for brainstorming, DALL·E for image generation | Ensemble of prompts: Gemini for source harvesting, o3 for hypothesis testing and reasoning, Grok for narrative writing | GitHub Copilot Agent + Sonnet; thought for 5 days, built in 4 |
For the purposes of this argument let’s just state that all three solutions were outstanding. This way we remove any bias due to idea variance.
Now, I want you to take a moment and put yourself in the shoes of an executive. One who has to convince other executives at the company about this opportunity. Which output empowers the executive the most?
Person A’s output is locked in Figma. The static screens likely need to be thrown into a PowerPoint, and with a little back and forth, the executive has essentially a static story that leaves everything to interpretation by the audience.
Person B’s output is a strategic document. The main empowerment this gives the executive is the data points they need to make a decision. Executives are key decision makers that weigh risks through data. So this format gives them the most aligned output to their natural way of working.
Person C’s output is a live demo, shared with a link to experience, feel, and potentially share with other executives in a more engaging way. Imagine the kinds of conversations this executive will have humble bragging about the work their teams are doing.
Now let’s ask another hypothetical question. A few months later, markets are tough and layoffs are everywhere. Cuts have to be made. In that kind of environment, who would the executive retain and who would they part ways with?
Person A isn’t keeping up with the times. Likely needs a lot of retraining. Person B gets a pass. Person C remains because they empower the executive beyond what they imagined was possible.
Business has ALWAYS been about being able to make high leverage strategic decisions and executing them at speed. AI enables everyone to be more productive beyond what was ever possible. But it now requires us to reframe the way we work.
Person A isn’t necessarily lacking talent. But just hasn’t developed the skills beyond rudimentary ChatGPT usage. Person B deeply understands which model is great at what and has made AI do the custom heavy lifting. Using AI to augment the strategy in ways not possible before and would likely take several weeks to research. Person C bends time by building the closest thing to the idea. The command of leveraging AI to get things done is diverse resulting in a superior outcome and possible trust.
That, in miniature, is the story of AI literacy in 2025.
The Mirage of Mass Adoption
Scroll LinkedIn for five minutes and you’d think the entire knowledge‑workforce has morphed into AI Researchers and Experts. The reality however is far from it.
Microsoft’s latest Work Trend Index surveyed 31,000 workers across 31 countries:
75% use gen‑AI weekly, yet only 40 % feel competent.
Executives, meanwhile, remain sanguine. In a McKinsey study, 99% of the C‑suite claimed they understood generative AI.
13% of employees say they already use generative AI for ≥30% of their daily work
92% of companies plan to increase AI spending over the next 3 years. Yet, only 1% consider themselves AI‑mature
A striking 47% of C‑suite leaders say their organizations are moving too slowly in developing and deploying AI, citing talent gaps as the main issue
What this tells us is despite all the AI hype you see on LinkedIn daily, organizations haven't quite figured out how to win with AI. All while doing little to help their own employees keep up with AI.
Despite experimentation, >80% of companies report no meaningful impact on enterprise‑wide EBIT
41% of employees are apprehensive, indicating a need for structured training and support
The Literacy Ladder
Typing once set executives apart from the secretarial pool, and spreadsheets later distinguished the data‑savvy from the rest. Today, the quiet but decisive marker is AI literacy. It can be mapped across three rungs.
| Category | Definition | Usage | Mindset | Tools |
|---|---|---|---|---|
| AI Illiterate | Surface-level users | Summarization, writing emails, light ideation, meeting notes | AI as a fancier search engine or grammar tool | ChatGPT, Copilot, Notion AI — default settings, little customization |
| AI Native | Proficient integrators who leverage AI to develop custom workflows | Automating tasks, prompt-chaining, tool orchestration | AI as a collaborator that increases leverage | GPTs, Code Interpreter, Lovable, Bolt, Replit |
| AI Creationist | Frontier-builders creating novel capabilities or primitives | Inventing tools, fine-tuning models, core science applications | AI as a new medium of invention | LangChain, Cursor, Copilot, PyTorch, Hugging Face, vector DBs, custom agents |
Each rung multiplies leverage, and leverage, history shows, calcifies into wealth.
The New Gilded Age of Skill
PwC’s 2025 AI Jobs Barometer quantifies the chasm: roles demanding AI fluency now command a 56% wage premium, up from 25 % a year ago. Industries most exposed to AI are booking revenue‑per‑employee growth triple that of laggards. Skills in those roles decay faster too—66% quicker turnover—dramatically widening the moat around the fluent.
Salary bands are shifting faster than anyone could imagine. In earnings calls, “model leverage” now sits beside revenue and margin—an urgent signal that value is migrating to those who wield AI with intent. Titles and tenure still matter, but they now trail a more decisive metric: how skillfully you bend a model to your domain. The early evidence is most visible in the middle rungs—product managers, analysts, including unexpected roles like warehouse supervisors—who turn bespoke AI workflows into exponential throughput, leaving prompt‑only peers behind.
Star Researchers Become Franchise Players
Meta’s new super‑intelligence lab reportedly offers up to $300M over four years—more than $100M in year one—to lure marquee researchers. Similar nine‑figure packages surface at OpenAI and Anthropic. The Wall Street Journal dubs it an “exploding‑offer talent war,” the closest thing tech has seen to NBA free‑agency.
Boards stroke these checks because a single breakthrough model can re‑price an entire market. But this is the pyramid’s apex, not the whole story.
Middle Rungs Pull Away
Across 25 industries, listing even one AI skill on a résumé now increases salary offers by 56%. IDC pegs the return on gen‑AI spend at $3.70 for every $1, and roles augmented by AI are posting 4× faster revenue‑per‑employee growth than un‑augmented peers. Advantage compounds: early adopters lock in raises indexed to company growth; late adopters chase a moving target.
Structural Reallocation, Not Doom
The World Economic Forum forecasts 92M roles displaced by automation through 2030—but 170M created, a net gain of 78M. Crucially, 40% of employers say they will trim headcount where workers fail to upskill. The divide isn’t researcher‑versus‑rest; it’s adaptive versus static.
Takeaway: You don’t need a PhD or a $100M bonus, but you do need to find a way to go beyond basic ChatGPT prompts—or risk watching the ladder retract.
Aim to Live Above the Algorithm
Below the algorithm is where most of the world already lives. Gig‑workers wait for surge pricing; content creators are in a forever doom loop chasing clicks, likes, and fake hearts that will send them into oblivion the second they stop posting. The math hums overhead, invisible yet absolute, slicing margin for the platform and cents for the human. Generative AI threatens to extend that hierarchy into the very heart of knowledge work—unless we climb the ladder.
A Brief History of Moving Upstairs
1860s — Mechanical looms reward those who can service gears, not just weave thread.
1980s — Lotus 1‑2‑3 turns balance‑sheet clerks into obsolete columns; MBAs who master macros double their bonus pool.
2020s — AI turbocharges anyone who can automate workflows, chain and fine‑tune models; those who refuse become spectators within a AI-enabled acceleration of the enterprise.
The pattern is cruel but clear: the first movers who author the automation capture its rents. Everyone else, meanwhile, supplies training data.
Three Ladders Out of the Basement
Agency — Edit the Algorithm
Ownership begins the moment you stop accepting default settings. A supply‑chain analyst who builds a private GPT agent to reconcile invoices isn’t just faster—she redefines what “reconciled” means in her firm. Decision‑rights shift with the code.
Adaptation — Learn at the Speed of Models
PwC’s 2025 data show skill requirements in AI‑exposed roles turning over 66% faster than in low‑exposure jobs. The safest resume is one written in pencil and updated quarterly. You have to keep learning aggressively.
Meaning — Ask the Non‑Obvious Question
Herbert Simon warned that wealth flows to those who manage attention. AI can generate articles, melodies, even legal briefs—but it cannot yet decide which problems are worth 10,000 GPU‑hours. Curiosity remains a moat.
The Moral Geometry of the Divide
If history holds, value will pool around the people who shape models, not merely invoke them. That does not require a PhD in transformer math; it does require an apprenticeship mindset—treating every workflow as provisional, every prompt as a prototype. The alternative is living below the algorithm: your pay pegged to a dashboard you didn’t design, your upside capped by someone else’s heuristic.
The climb from consumer to author is short, but the drop from author to obsolete is shorter still.
