According to Forbes, based on a new Google paper and interviews with global leaders, AI is already transforming how people learn and work in tangible ways, from a teenager in Kenya using an AI coding tutor to conservationists in Papua New Guinea analyzing field data. The article cites McKinsey Global Institute research warning of a historic workforce transition, with many current tasks having “technical potential” for automation. Key figures like UN/IOE Ambassador Shea Gopaul, Google’s Ben Gomes, and DeepMind COO Lila Ibrahim argue the outcome depends on building systems that close access gaps for the 2.6 billion people still offline and focus on employability. They highlight pilots like Google’s LearnLM, designed with neuroscientists to act as a tutor, and shorter, stackable apprenticeships that can last just 3-6 months, aiming to directly connect training to hiring.
The reality check before the revolution
Here’s the thing that often gets lost in the shiny AI hype: the ground floor is missing for a huge chunk of the planet. Ambassador Shea Gopaul’s point is brutal and necessary. When we talk about AI tutors and digital apprenticeships, we’re assuming a baseline of electricity, internet, and basic digital literacy that simply doesn’t exist for about a third of the world. And the people in the most precarious jobs—the informal economy that employs up to 90% of workers in some regions—are the ones who could benefit most from upskilling, but have the least access to it. It’s a classic case of the Matthew effect: those who already have, get more. So the big, unsexy challenge isn‘t just building smarter AI. It’s building the literal and figurative infrastructure so that AI doesn’t become another force for inequality. Fail to plan for that, and all the cool pilots are just that—pilots.
From learning to earning, faster
This is where the Google perspective gets interesting. Ben Gomes pushes back on the fear that AI makes learners lazy. I think he’s onto something. The real enemy in education isn’t asking for help; it’s “metacognitive laziness”—spinning your wheels on poorly presented material or getting stuck in guesswork. If AI can strip that away and let a student focus on the core conceptual struggle, that’s a win. But the more compelling shift is directly tying learning to job pipelines. Using AI-powered lab simulations as the first stage of recruitment? That’s a game-changer. It moves the credential from a piece of paper to a demonstrable, hands-on skill. Suddenly, the training isn’t abstract. It’s a direct audition. This is the pattern that matters: tools that save educators time and create a visible, credible bridge to a paycheck.
The engine under the hood: trust
Lila Ibrahim’s description of LearnLM is crucial because it highlights the difference between AI for learning and AI that’s just near learning. Throwing a powerful LLM at a student and calling it a tutor is a recipe for disaster—it’ll just hand over answers. Building a model from the ground up with cognitive scientists and educators, one trained to guide and ask questions? That’s a different beast. It acknowledges that pedagogy is a discipline, not a data stream. For anyone evaluating these tools—schools, parents, businesses—this is the litmus test. Was this built with learning science, or just repurposed? Does it explain, or just output? The safety and bias monitoring for a learning model also needs to be orders of magnitude stricter than for a generic chatbot. Get this engine wrong, and the whole vehicle goes off a cliff.
The view from the ground
Maybe the most important voices here are the young innovators from the Global South. Enzo Romero and Dikatauna Kwa aren’t talking about theoretical job markets. They’re describing AI as a practical lever right now: to design prosthetics, analyze coral data, write grant proposals. Their point is devastatingly simple: AI can democratize high-skill work, but only if it’s accessible and context-aware. An AI tutor trained on a New York curriculum is useless in rural Peru or Papua New Guinea. So their ask isn’t for more hype. It’s for reliable internet, localized models, and infrastructure so their communities aren’t permanently in “pilot mode.” They’re already building the future. The question is whether the rest of the world will build the platforms that let them scale it.
So who actually wins?
All this circles back to Gopaul’s final, hard-nosed point about governance. Cool tools are great. Pilots are fun. But without clear ownership, accountability, and a governance blueprint, this all becomes a risky experiment on the global workforce. The McKinsey research shows the potential disruption is massive. The Google paper outlines a potential path. But the roadmap they all seem to agree on is pragmatic: invest in access, build tools for understanding (not just answering), and tether every training program to a real job outcome. Basically, measure success not by how many people get a digital badge, but by how many get a decent job. It’s not a tech problem anymore. It’s a political and organizational one. And that’s always been the harder puzzle to solve.
