Most businesses haven’t made that change yet. They still think of learning technology as a purchase decision. They buy an LMS, add some AI tools, and call it a learning stack. The result looks modern on a slide, but it doesn’t work as well in real life. The delivery of content gets better. But skills development doesn’t.
The gap is structural. A real learning infrastructure has four pillars, and missing any one of them causes the whole system to fall short. National-scale strategies recognise this. Most enterprise L&D functions are still catching up.
Pillar 1: Compute
Compute is the foundation that everything else sits on. It is the processing capacity that allows AI models to run, content to be generated, and learning interactions to be delivered in real time.
For most enterprises, compute is invisible. It is something the IT team handles, abstracted behind cloud services and vendor platforms. That works at small scale. It breaks down when learning becomes continuous and AI-driven, because the volume and complexity of interactions grow faster than legacy procurement assumptions accounted for.
National AI+education strategies treat compute as a strategic asset. They invest in dedicated infrastructure designed to support educational workloads at scale. Enterprises that want to operate learning the same way need to make similar architectural decisions. That does not mean building data centres. It means understanding which workloads need dedicated capacity, which can run on shared infrastructure, and how compute costs will scale as AI usage grows.
Pillar 2: Data
Data is what makes learning intelligent. Without high-quality, well-structured data on learners, content, performance, and outcomes, even the most advanced AI models produce generic results.
According to the IBM Global AI Adoption Index, the largest barrier to AI deployment in enterprise settings is not the technology itself. It is the readiness of the underlying data. The same is true in learning. Most organisations have learner data scattered across HR systems, LMS platforms, performance review tools, and individual department spreadsheets. None of it talks to each other.
A national AI+education strategy sees data as one resource. Skills data, assessment data, performance data, and demographic data are linked so that a full picture can be used to plan learning interventions. Companies that want the same level of capability need to do the same work in-house. Before they buy more AI tools, they need to put money into data integration. Adding intelligence to data that is already broken up only gives them broken insights.
Pillar 3: Models
Models are where most of the recent excitement in learning technology has been concentrated. Generative AI, adaptive learning engines, skills inference algorithms, and personalisations systems all sit at this layer.
But models are only as good as the compute and data they sit on. A sophisticated personalisations engine running on incomplete data produces confident-sounding recommendations that miss the mark. A generative AI tool with no connection to enterprise context produces generic content that learners ignore.
The lesson from national-scale strategies is that models should be selected and deployed based on the specific outcomes the system is trying to produce, not on novelty or vendor pressure. The right model for compliance training is not the right model for leadership development. The right model for skills assessment is not the right model for content generation. Enterprises that treat models as a portfolio matched to specific outcomes consistently outperform those that buy whatever is currently being marketed most aggressively.
Pillar 4: Outcomes
Outcomes are the pillar most often missing from enterprise learning architectures. Compute, data, and models can all be in place, and the system can still fail to deliver measurable business value if outcomes are not defined upfront and tracked rigorously.
LinkedIn’s 2025 Workplace Learning Report found that only a minority of L&D functions can directly tie learning activity to business performance. The rest are measuring outputs such as courses completed and hours consumed, rather than outcomes such as skills acquired, performance improved, and business problems solved.
National AI+education strategies start with outcomes. Workforce readiness, economic competitiveness, and educational equity are defined as the goals, and the rest of the stack is designed to produce them. Enterprise L&D leaders need to apply the same logic. The question is not what learning technology to buy. The question is what business outcome the learning system is built to produce, and whether the four pillars are configured to produce it.
Why Missing One Pillar Breaks the Whole Stack
Each pillar depends on the others. Compute without data produces nothing useful. Data without models produces dashboards no one acts on. Models without outcomes produce activity that does not move the business forward. Outcomes without the underlying infrastructure produce frustration, because the goals are real but the system cannot deliver against them.
The reason national AI+education strategies are worth studying is not that they are perfect. It is that they take infrastructure seriously. They make the architectural decisions enterprises tend to skip. And they invest in all four pillars at the same time, because partial investment produces partial results.
Building a Learning Infrastructure That Holds
Most enterprises do not need to operate at national scale. But they do need to think structurally. A learning function that wants to deliver real outcomes in 2026 and beyond needs all four pillars in place, working together, and aligned to business priorities.
If your organisation is rethinking its learning infrastructure, we can help you build a stack that delivers.
Talk to our team at zillearn.com/contact-us/
Sources: China Ministry of Education. “Action Plan for AI+Education.” (April 2026) IBM. “Global AI Adoption Index.” https://www.ibm.com/think/insights/ai-adoption LinkedIn. “2025 Workplace Learning Report.” https://learning.linkedin.com/resources/workplace-learning-report