Understanding the Skills Gap in AI Workforce

AI Literacy or AI Elitism? The Hidden Divide in Workforce Readiness The OECD - OCDE report "Bridging the AI Skills Gap: Is Training Keeping Up?" (2025) investigates whether current training systems are adequately equipping workers with the skills needed for an AI-driven economy. It documents experiences from 21 OECD countries, including detailed analyses of training policies and the actual content of course catalogues in Australia, Germany, Singapore, and the United States. The report finds that although most countries are increasingly promoting upskilling and reskilling for AI, only a tiny share of courses (between 0.3% and 5.5% in the four sample countries) cover AI content, with a disproportionate focus on advanced skills rather than general AI literacy for the wider workforce. #KeyProblems identified include insufficient supply of AI training, over-reliance on financial incentives that are not always targeted to AI skills, and a lack of flexible, inclusive pathways—especially for vulnerable or less-skilled workers. Experiences from various countries reveal that while some are pioneering targeted AI literacy campaigns (e.g., Austria’s “Digital Everywhere” and Hungary’s gamified AI challenge), the majority remain focused on producing AI professionals. #Recommendations #for #policymakers are to expand both general and advanced AI training, better target incentives, increase non-financial support (like career guidance and public-private collaborations), develop more accessible and flexible training formats, lower entry barriers for participants, and embed AI skill-building within broader workforce development strategies. These steps are crucial to ensuring a people-centred, inclusive transition as AI reshapes the world of work. Source: https://lnkd.in/eMTMPQSy

The gap I am focused on most these days when it comes to AI at work, is the gap between employees and employers. We know that 75% of knowledge workers are using GAI on the job, saying it’s not just helping them save time to focus on more important work but also to bring more human skills to their work, like creativity. But we also know that only 39% of those workers have been trained on AI at work, as companies struggle still to come up with a point of view on AI as well as a strategy for workforce development in the age of AI. If your company is struggling on that part, one thing you can do is look to those who are leading the way. IBM and Siemens are great examples of companies who are two steps ahead of most, moving beyond the incremental early days of AI towards the real, transformative benefits. I was inspired by my conversation a few weeks ago with Nickle LaMoreaux and Brenda Discher who are not only innovating with AI at scale, but keeping people at the center of it all. Across those conversations and many others I’m having, a few key foundational steps are emerging: 1️⃣ Have a pro-human AI point of view and strategy in place. AI has the potential to build a world of work where people can bring their full skills and abilities to bear — but we need to believe in the power of our people more than the power of our tech to realize it. 2️⃣ See jobs as tasks, not titles. Once you boil down a job down into a set of tasks, it’s much easier to see where AI is coming in to change or disrupt some of those tasks and where there are uniquely human skills people will spend much more time on then before. In a world where 68% of skills are set to change by 2030, understanding where this change will hit is crucial to helping your teams stay resilient. 3️⃣ Build learning into the day to day of your company’s culture. As skills for jobs change rapidly – learning is no longer a one-off moment at the start of a career. The ability to learn, unlearn, and relearn is what sets teams apart to stay agile and resilient. 

There’s a common belief that AI will close the skill gap between beginners and experts. But a new Stanford-Harvard study shows it’s more complicated. Researchers studied three groups inside an organization, all asked to write web articles: • Insiders: SEO web analysts who regularly wrote articles • Adjacent outsiders: marketers in the same department who didn’t usually write • Distant outsiders: technologists whose work was unrelated to writing AI didn’t boost every group equally. It helped adjacent outsiders close the gap and perform at insider level. But distant outsiders (technologists) still fell short. Even with AI fine-tuned on company documents, their articles scored lower than insiders’. They hit what the researchers call the “GenAI wall.” They lacked too much of the marketers’ tacit knowledge: the instinct for tone, the craft of building a narrative, the ability to weave ideas coherently, and the judgment of what makes an article “good.” AI couldn’t fill that gap or replicate the intuition that insiders had built through experience. One of the next big challenges, and opportunities, for AI is learning tacit knowledge. Bridging the gap between novices and experts requires more than data. It means capturing the shortcuts, sequences, and judgment calls that turn a draft into a finished product. Only when AI understands the real workflows, rhythms, and processes of an organization can it start to absorb that hidden know-how—and begin to shift the “GenAI wall.”

Most companies I speak to are quietly anxious about AI. Not because they don’t know what AI is. But because they don’t know how their teams are actually using it. A few people are experimenting. A few are secretly very good. Most are stuck copying prompts from instagram and hoping it helps. Leadership thinks, “We’ve rolled out ChatGPT access, that should be enough.” It isn’t. The real gap is not tools. It’s workforce readiness. Who in your team: >> Knows how to use AI beyond writing emails? >> Can redesign their own workflows using AI? >> Is confident enough to rely on AI for decisions, research, and planning? Most companies don’t have answers to this. They only find out when execution slows down or competitors move faster. This is why “AI upskilling” cannot be a one-day workshop or a generic course. It needs: >> Benchmarking, so you know where your workforce actually stands >> Role-based upskilling, not one-size-fits-all sessions >> Workflow redesign, so AI shows up inside real work >> Ongoing support, so adoption doesn’t drop after excitement fades If your team feels overwhelmed, confused, or uneven in AI usage, that’s normal. What’s risky is ignoring it. We’ve been helping leadership teams and workforces move from AI curiosity to AI-powered execution. If this is something you’re thinking about for your company, lets talk!

The most important skill we’re not measuring yet? AI literacy. And no, it’s not just asking ChatGPT to make something sound better. It’s about knowing how to work with AI—when to trust it, when to question it and how to lead alongside it. That takes more than technical know-how. It takes critical thinking, adaptability, and real world experience. Skills that are deeply human. But here’s the gap I’m seeing –  we’re moving fast on tools, but not nearly as fast on the skills people need to use them well. Are we teaching people how to work with AI? Are we measuring what they actually understand? How do we know it’s working if we’re not paying attention to what people are learning? If AI is part of how we work, AI literacy needs to be part of how we lead. #AI #AILiteracy #FutureofWork

𝗧𝗵𝗲 𝗡𝗲𝘄 𝗦𝗸𝗶𝗹𝗹 𝗚𝗮𝗽 𝗶𝗻 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗔𝗜: 𝗣𝗿𝗼𝗺𝗽𝘁𝗶𝗻𝗴? 𝗼𝗿 𝗦𝗰𝗼𝗽𝗶𝗻𝗴? There is a growing belief that if teams master “prompt engineering,” they will somehow become AI-ready. So enterprises run workshops, circulate prompt patterns, publish cheat sheets, and certify employees. After all that effort, the business still does not see meaningful outcomes. Productivity bumps are anecdotal, KPIs stay flat, and leaders wonder why AI is not delivering the transformation they were promised. Because prompting was never the real skill gap. The biggest blockers appear long before anyone writes a prompt. It starts with problem framing. You can build an “AI summarizer” for a 200-page compliance document, but after a few conversations the real objective emerges: reduce onboarding time by 40 percent. That requires workflow redesign, not clever prompting. Clear framing would have exposed that immediately. The next wall is data availability. And the surprise is always the same: having data and having usable data are completely different. In most enterprises, scoping becomes a negotiation with data owners, compliance teams, and buried spreadsheets. The real questions are about completeness, quality, structure, accessibility, and policy constraints. Teams keep tuning prompts when the deeper issue is that the model cannot learn from what it cannot legally or reliably access. The blocker is the organisation, not the LLM. Then there is the economics of AI. I covered this in an earlier post, but the summary is simple: AI is not automatically the smartest or cheapest option. Many workflows can be improved with simpler automation, and an LLM is worth using only when the business impact justifies the cost. Finally comes change management, the quiet breaker of AI projects. Even when the model works and data pipelines are clean, adoption fails because people are not ready for the new workflow. Ownership shifts, trust dips, and middle managers quietly resist. Every AI project has two launches: the technical go-live and the human go-live. The second one decides everything. And none of these challenges get fixed by better prompts. Prompting can polish a solution, but it cannot rescue a poorly scoped one. It does not fix data gaps, align incentives, modernize workflows, or overcome cultural resistance. Prompting is tactical. Enterprise AI is strategic. This is why scoping is where real AI maturity develops. Scoping clarifies what problem is worth solving, whether the data is usable, what the economics look like, and what human changes are needed for adoption. At its core, scoping is about asking the right questions before building anything. What outcome are we trying to change? What data do we truly have? What will this cost to operate? How will people use it day to day? These questions separate mature AI organizations from those that experiment with prompts and hope for the best. Cognida.ai #PracticalAI #EnterpriseAI

These headlines all say the same thing: "Train more people on AI tools." But they're all solving the wrong problem. Here's what I've seen running AI training with founders over the last year: The gap isn't "How do I use ChatGPT?" The gap is "How do I THINK about my business problems so AI can actually help?" I've watched companies spend lakhs on tool training. Teams learn prompts, shortcuts, fancy workflows. Three months later? Back to doing things the old way. Because nobody taught them to ask better questions. Nobody helped them unbundle their job into "tasks AI should do" and "tasks only I can do." The headlines have it backwards. India doesn't have an AI skill gap. India has an AI thinking gap. The companies that will win by 2027 aren't the ones with the most AI-certified employees. They're the ones whose people deeply understand their customers, their problems, and their craft — and know WHAT to ask AI. WHO not HOW. Always. What's the bigger gap in your organization — AI skills or AI thinking? #CyborgMindset #AIAdoption #FutureOfWork #AITransformation

I still remember the face of the senior developer who walked into my Lahore office last year, résumé in hand, asking if his 10 years of experience still mattered. "Mr. Usman," he said, "I built systems that run dozens of Technology and retail organizations, but today a 22 year old with three months of AI training just got promoted over me." That moment haunts me, because it represents the greatest challenge facing every technology leader today. The World Economic Forum estimates that nearly six in ten workers will require training before 2030, with 22% of jobs globally changing due to technological advancements. Yet 80% of organizations say upskilling is the most effective way to reduce employee skills gaps, but only 28% are planning to invest in upskilling programs over the next two to three years. This disconnect isn't just a statistic; it's a crisis of vision. Gartner projects that generative AI will require 80% of the engineering workforce to upskill through 2027, while 60% of employees report insufficient training for core job skills. We're essentially asking people to swim while refusing to teach them. But here's what I've learned building Devsinc across three continents: the solution isn't more training programs. It's about creating what Deloitte calls a "whole work approach to development" that integrates skill building with practical, contextual experience. When you redesign roles and workflows to reduce reliance on missing skills while providing intensive, hands-on management support, you don't just close gaps; you create cultures of continuous learning. That senior developer? He's now leading an AI integration team. Not because we gave him a course, but because we paired him with younger engineers in a true knowledge exchange. His decades of system thinking combined with their AI fluency created something neither could achieve alone. 46% of leaders identify skill gaps as the most significant barrier to AI adoption. But the real barrier is our failure to see that experience and innovation aren't competitors. They're collaborators waiting to be unleashed. The question isn't whether your people can adapt. It's whether you're brave enough to invest in their transformation before your competition does.

Most companies wait until they have an urgent problem before addressing workforce capability. But the ones building competitive advantage are investing in readiness before the gap becomes a crisis. Here are four areas where organizations need to focus: 𝟭. 𝗥𝗲𝘀𝗸𝗶𝗹𝗹𝗶𝗻𝗴 𝗳𝗼𝗿 𝗿𝗼𝗹𝗲𝘀 𝘁𝗵𝗮𝘁 𝗱𝗶𝗱𝗻'𝘁 𝗲𝘅𝗶𝘀𝘁 𝗳𝗶𝘃𝗲 𝘆𝗲𝗮𝗿𝘀 𝗮𝗴𝗼 Automation specialists, data scientists, and AI integration roles require new training pathways. Companies that build apprenticeship programs and internal development tracks get ahead of skills bottlenecks before they slow growth. 𝟮. 𝗣𝗿𝗲𝗽𝗮𝗿𝗶𝗻𝗴 𝘁𝗲𝗮𝗺𝘀 𝘁𝗼 𝘄𝗼𝗿𝗸 𝗮𝗹𝗼𝗻𝗴𝘀𝗶𝗱𝗲 𝗔𝗜 It's not enough to deploy AI tools. Teams need to understand how to integrate AI into their workflows, manage AI-driven processes, and improve performance through human-AI collaboration. 𝟯. 𝗜𝗱𝗲𝗻𝘁𝗶𝗳𝘆𝗶𝗻𝗴 𝘀𝗸𝗶𝗹𝗹 𝗴𝗮𝗽𝘀 𝗯𝗲𝗳𝗼𝗿𝗲 𝘁𝗵𝗲𝘆 𝗮𝗳𝗳𝗲𝗰𝘁 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 Skills assessments show what people can actually do, not just what their job titles suggest. Companies that map capabilities across their workforce can redeploy talent strategically and keep people engaged in roles where they can grow. 𝟰. 𝗖𝗿𝗲𝗮𝘁𝗶𝗻𝗴 𝗽𝗮𝘁𝗵𝘄𝗮𝘆𝘀 𝗶𝗻𝘁𝗼 𝗿𝗼𝗹𝗲𝘀 𝘄𝗵𝗲𝗿𝗲 𝗽𝗲𝗼𝗽𝗹𝗲 𝗰𝗮𝗻 𝘀𝘂𝗰𝗰𝗲𝗲𝗱 Whether it's technical training, role-specific development, or management skills, companies need structured programs that prepare people for the work that's coming, not just the work that exists today. The retirement wave is gathering speed. Skills-based hiring is becoming the norm. Growth isn't waiting. What's your approach to workforce readiness right now?

I’ve been thinking about why the companies with the most AI training hours are often the ones with the least to show for it. They bought the seats, ran the workshops, hired the consultants. The problem isn’t adoption. It’s that the skill the economy actually needs doesn’t have a name. It doesn’t have a curriculum. And it doesn’t work like any workforce skill that’s come before it. Picture a bubble. The air inside is what AI agents can do reliably. The air outside is what still requires a human. There's a thin surface between the two. That surface is where the interesting work happens: deciding what to delegate, how to verify, where to intervene, when to trust. It's where the value concentrates. When the bubble inflates, the surface area increases. Every capability jump creates more boundary to operate at, not less. More seams, more judgment calls, more decisions about where human attention creates value. The infrastructure we’ve built to teach workforce skills assumes the target stands still. Every prior workforce skill — literacy, numeracy, computer literacy, coding — was a destination. You reached it, you had it, you moved on. But the skill of working at the surface of this bubble has no fixed destination, because the surface keeps expanding outward. I think that mismatch is the most expensive gap in the global workforce right now. In November 2025, the state of the art was one thing. By February 2026, it was a generational leap compressed into a single quarter. The person who started building calibration in November has three months of updated operational intuition that the person starting now doesn’t have. If you set policy: Stop building curricula. Start building flight simulators. The workforce development infrastructure for the next ten years is practice environments that expose workers to realistic agent capabilities, realistic failure modes, and realistic conditions that change. Wrapping up this mini brief with an audit question to run this week: Can you name, specifically, three tasks your team delegated to agents last quarter that they didn’t delegate the quarter before? -- Every Sunday I write a briefing for our Executive Circle. We launched a WhatsApp group a couple of weeks ago. It's already one of my favorite places to spend time, with senior leaders working through the same strategic decisions. If you’re a member, come join the conversation. Access here: https://lnkd.in/e4UBrRcV