AI is accelerating the pace at which products evolve, workflows change, and employees need to refresh their skills. In this environment, many organizations are discovering that their traditional training approaches of long courses, slow development cycles, and centrally controlled content can no longer keep pace with the realities of work.
Across the companies I speak with — from large enterprises to individual creators — a common pattern emerges. The challenge is not usually the AI tools themselves. It is the growing gap between how quickly work changes and how slowly many organizations can update and distribute knowledge. Institutional knowledge now decays faster than traditional training can keep up.
This is why I encourage leaders to think of learning not as a standalone program but as a Learning Operating System where knowledge runs continuously in the background, is updated in real time, and can be easily accessed by everyone. In practical terms, this requires three key shifts.
From static courses to continuous, contextual learning
For a lot of companies, learning is still a static, bureaucratic process. That model is completely inadequate in a world where knowledge is created and quickly becomes obsolete. That’s not how people learn.
I often tell teams that people simply can’t spare 45 minutes for a training session. Give them searchable micro-resources that live inside the tools they already use: their CRM, helpdesk, Slack, and code repos. Use AI to surface the next best learning step based on what someone’s actually doing right now. And make application mandatory: job aids, checklists, simulations with real scenarios. If they can’t use it immediately, don’t build it.
This is where AI can help transform raw knowledge into flexible learning assets. Learning has to be continuous and embedded in work. It has to be tightly integrated with daily workflows.
This is learning in the flow of work in practice — knowledge that appears exactly when someone needs it.
From centralized production to decentralized creation with shared standards
In many organizations, training is still produced by a small central team. But the subject-matter experts closest to the work should contribute, as their knowledge changes the fastest.
AI finally makes it possible for anyone to create meaningful training without a background in instructional design. It can take raw expertise and structure it into lesson plans, suggest outlines, generate examples, and turn knowledge dumps into questions or scenarios.
From what we’re observing in our own use of GenAI, the bottleneck around ‘who can teach’ is disappearing. Experts with no design background can now turn their knowledge into structured learning, which fundamentally broadens who contributes to training.
In this new world, the role of L&D becomes more strategic; L&D becomes the editor and architect, responsible for standards, coherence, governance, and instructional quality. This includes ensuring one authoritative version of each important lesson or process, and providing lightweight human review so AI-generated content stays accurate, relevant and in the right tone. In short: Decentralize creation, centralize standards. This shift ensures knowledge stays current, coherent and accessible.
From “more content” to learning that is measured like a product
When AI can generate content in seconds, the meaningful question is no longer “Do we have a course?” but “Does this learning actually change performance?” This requires treating learning like a product with clear owners and business metrics. I often advise clients to assign an owner for each critical learning journey (onboarding, manager development, product enablement). And define success in business terms, such as ramp, time-to-productivity, ticket deflection, and talent retention.
-
Onboarding & ramp: Track time-to-productivity, time-to-first meaningful outcome, and whether new hires are making fewer mistakes.
-
Customer & product education: Measure product activation and adoption, reduction in repeated ‘how do I’ questions, and renewal and upgrade rates for accounts with trained users versus those without.
-
Internal skills development: Look for reduction in specific errors or incidents, and internal mobility and promotion rates for people who follow defined learning paths.
-
Health of the learning system: Measure frequency of updates, and participation and completion in the most important pathways, not just total logins.
This moves learning from counting completions to demonstrating real business impact.
Turning scattered training to a functioning Learning OS
I see a common pattern, especially in mid-sized organizations. The initial state is where training is scattered across slide decks, PDFs, and ad-hoc sessions. Onboarding feels long, inconsistent, and heavily dependent on which manager you get. Knowledge lives in the heads of a few senior employees. That changes when leaders decide to treat learning as a growth lever, not a cost center, by creating an internal academy or school. They capture the practices of their best performers through recordings, playbooks, and examples, and use AI to create structured paths, assessments, and practice activities for each role.
Within months, improvements can be seen:
-
New hires reach basic productivity faster because they follow a clear path instead of hunting for information.
-
Managers report spending less time reteaching fundamentals and more time coaching higher-value skills.
-
Customer experience becomes more consistent because teams learn from the same source of truth.
Working across both creator-led learning and organizational training, we’ve seen how the expectations shaped by passionate creators now spill into the workplace. People increasingly look for the same kind of engaging, consumer-grade learning inside their companies as well.
A Learning Operating System is what lets organizations meet those consumer-grade expectations: it makes knowledge easy to update, easy to contribute to, and easy to access in the moment of need — decentralized, fluid, and woven into daily work.
4 actions HR and L&D leaders can take in 2026
-
Start from one shared problem and metric per function. Learning becomes far more effective when it is anchored to the real work of each team. With Sales, that might be improving discovery calls or shortening ramp, measured by time-to-productivity or win rate on a specific product. With Product, it might be reducing confusion after a release, measured by feature activation and usage.
With Customer Success or Support, it could be reducing a cluster of repeated “how do I” tickets, measured by volume, deflection, and CSAT. Starting from a shared problem gives learning a clear target and replaces vague goals with measurable business outcomes.
-
Use real work as the raw material for learning. Most organizations already have the ingredients for great learning, they’re just hidden inside everyday work.
-
Sales: call recordings, proposals, win/loss notes
-
Product: internal demos, release notes, FAQs
-
Customer Support: recurring tickets and troubleshooting steps
-
AI makes it easy to turn these into short scenarios, micro-lessons, examples, and quick-reference guides, instead of reinventing content from scratch. This not only accelerates creation but also ensures learning matches the actual challenges people face.
-
Assign joint ownership and review metrics together. Every cross-functional learning stream should have one owner from the function (Sales, Product, CS) and one learning owner who acts as the architect. Their shared job is to keep the content current, review the agreed metrics, and decide what to change next. When you review ramp times, adoption patterns, error rates, or ticket trends together, something important happens: you adjust the learning, not just the messaging.
If numbers aren’t moving, it’s the scenarios, examples or practice that need updating—not the slide deck.
-
Surface learning inside the tools where work happens. While the academy or LMS remains the structured backbone, the most-used learning assets are the ones embedded directly in workflow- CRM, helpdesk systems, engineering tools and internal documentation. A short, vetted learning asset that answers the question in context always outperforms a 45-minute module. This is where AI personalization shines: surfacing the next best step based on what someone is actually doing.
Your Strategy Is Only as Strong as Your Learning
If I could leave CEOs with one insight, it would be this: Don’t treat learning as a support function; it’s the mechanism that multiplies every other investment in your business. If you invest in product but not in education, adoption will lag, and the value you hoped to unlock simply won’t be realized. If you invest in hiring but not in development, you’re renting capability instead of building it—and when people leave, all that expensive knowledge walks out the door. And if you invest in strategy but not in learning, you’re assuming your current capabilities are sufficient for your future goals. That has rarely been true, and it is even less true now in the era of AI.
It all comes back to a simple question: What are we trying to get better at as a company, and how deliberately are we learning it? The companies that answer that question honestly grow faster and are more sustainable. The ones that don’t are renting their future. They’re borrowing capabilities from whoever last trained their people, rather than building what they’ll need next.
