Why your AI L&D strategy needs roots first
Over the past two years, I have been in continuous dialogue with CHROs, CLOs, and heads of digital learning at enterprises, digital publishers, and learning technology platforms. Almost every organization is moving forward with AI learning initiatives. Investments are being made, pilots are underway, and expectations from executives are high.
But once these conversations go beyond the surface, a consistent and uncomfortable pattern emerges.
Despite AI investment in L&D reaching record levels, measurable impact on employee performance remains elusive. Content creation is faster, but application is not. The pilot appears on the dashboard, but it doesn’t scale. And the skills gap that executives most urgently want to close remains large.
According to BCG, 74% of organizations report seeing no tangible business value from their AI investments, despite totaling $252.3 billion in AI spending in 2024 alone. MIT’s 2025 GenAI in Business study found that 95% of GenAI pilots fail to demonstrate impact on the bottom line, and S&P Global reports that 42% of companies will abandon most of their AI initiatives in 2025, up from just 17% the year before.
When it comes to learning specifically, LinkedIn’s 2025 Workplace Learning Report reports that while 80% of L&D professionals believe AI is important to their strategy, only 25% are incorporating AI into their day-to-day work. Meanwhile, 49% of learning and talent professionals say executives are concerned that employees don’t have the right skills to execute business strategy.
This is the AI learning gap that no one is talking about too loudly: the gap between investment and actual workforce capabilities.
In my view, and in Harbinger’s work supporting the world’s leading digital publishers, associations, and enterprise learning teams, the root cause is not technology. This is the foundation on which AI learning strategies are built.
AI in learning is a system shift, not a tool upgrade
The most common starting point I see is organizations treating AI as a way to do things they were already doing faster, such as building courses faster, generating assessments at scale, and automating translation and localization. These are true efficiencies. However, the learning mechanism remains the same.
AI fundamentally changes the economics of learning content. What used to take 40 hours now takes 4 hours. But if your content remains in a SCORM package where no one opens past slide 12, you’ve just created a mediocre artifact faster. Learner expectations are also changing. People want support built into context-sensitive, just-in-time work flows, not courses launched from an LMS.
This creates structural demands on the learning ecosystem that most organizations have yet to meet. Content can no longer be static. Systems must continually evolve. The underlying architecture must support module reuse, AI interactions, and context distribution across channels.
When organizations overlay AI on top of traditional course-centric models without addressing these structural realities, the results are predictable. AI will not change a broken system. It reveals and accelerates its limits.
Where most AI learning strategies fail
At Harbinger, we consistently see the same pattern of failure when working with companies and transforming digital publishing.
Content isn’t ready: Most learning ecosystems are built on SCORM packages, PDFs, and linear video—formats designed for delivery, not machine interaction. Without structured metadata and modular architecture, AI systems lack the context they need to produce reliable output. As a result, more time is spent validating AI-generated content than reaping its benefits.
McKinsey’s 2025 State of AI report highlights that 51% of organizations have experienced at least one negative AI-related incident in the past year (most commonly output inaccuracies and non-compliance), making this a significant liability in the regulatory sector.
Treat modernization as a one-time project: Organizations begin content migration or platform upgrades and wait for the next budget cycle. In an AI-driven environment, content cannot remain static. Without a continuous modernization workflow, organizations find themselves forever behind.
Governance as an afterthought: AI enables speed. But speed brings risks if governance is not built in. Organizations are often hesitant to scale AI because they lack confidence in how errors will be detected, fixed, and audited.
Role ambiguity within the learning function: As AI enters the workflow, instructional designers, SMEs, and QA teams are unsure of how their work will evolve. This ambiguity creates friction and slows adoption, not because people are resistant to AI, but because no one is redesigning the operating model.
Being disconnected from business results: Perhaps the most significant failure. Most AI learning strategies are measured in terms of efficiency, such as time saved or course creation. Business leaders are now asking a different question. “Are employees actually more capable?” Are critical skills gaps closing? That question is difficult to answer honestly when learning remains focused on content creation rather than competency building.
What the evidence shows about highly mature organizations
LinkedIn’s 2025 Workplace Learning Report is informative. Only 36% of organizations are “career development champions” – those who systematically connect learning to career paths, internal mobility and business outcomes. However, in reality, companies are seeing measurably different results, including increased profitability, improved talent retention, and significantly higher AI adoption rates. Career development champions are 32% more likely to offer AI training and 51% more likely to consider themselves at the forefront of generative AI adoption, compared to just 36% of less mature organizations.
This pattern is consistent with what we see in Harbinger’s own offering efforts. In other words, the organizations that leverage AI the most are not the ones that started using the tools early. They are the ones who got the content infrastructure and operating model right the first time.
We present two examples from our work.
In one large course industrialization effort, similar to our work with healthcare and compliance content publishers, an organization had thousands of courses, each customized for different audiences. Rather than migrating the content as-is, it was decided to rebuild the content into reusable learning objects with appropriate metadata tagging. Since then, content production speed has increased 10x and automation rates have increased by 80%. More importantly, the modular structure allows you to update your content once and automatically republish it in a variety of formats. AI was the accelerator. Architecture was the foundation. (This reflects the work we have done for our customers in the healthcare and compliance training space, including our 6000 course automation effort in the clinical education space.)
In another case, a leadership development organization moved from a static course format to a structured, single-source content model. As content became modular and metadata-rich, AI-powered personalization became possible not because we adopted new tools, but because content finally became machine-readable. AI coaching simulations, dynamic assessments, and adaptive pathways are all now possible as downstream applications of the initial structural work done.
Pattern: System design occurs before capturing the value of AI.
Practice Model: Content Maturity × Operational Model Maturity
This helps you think about your AI learning strategy in terms of two dimensions: content maturity (how structured, modular, and reusable your content is) and operational model maturity (whether learning functions are performed in project-based workflows or continuous delivery).
Organizations with unstructured content and project-based workflows have found that AI creates more rework than value.
As content becomes more structured, it improves reuse and consistency…but unless the operating model changes, scale will remain limited. True transformation occurs when both aspects mature together. Mature organizations build modular content systems supported by continuous workflows and built-in governance. In such an environment, AI becomes a natural system extension rather than a bolt-on.
This dual maturity lens informs how Harbinger approaches AI-enabled conversations with clients. Whether our clients are enterprise L&D teams looking to transition from content distribution to workforce functions, or digital publishers looking to transform their catalog of PDFs into an AI-enabled content supply chain.
Differences in how highly mature teams work
The most sophisticated learning organizations I’ve worked with have one defining characteristic in common: they don’t start their AI journey with tools. It starts with system design.
They treat content not as a finished product, but as infrastructure. Content is divided into modular components, enriched with metadata, and designed for reuse. Courses, performance support tools, AI co-pilots, and analytics systems all come from the same source.
They reconsider their ratings. Instead of fixed, linear assessments embedded in courses, we build a dynamic system where questions are tagged by skill, complexity, and context. This allows assessments to be adapted based on learner responses and generates richer data on actual development, not just completion.
They don’t just readjust roles, they redesign them. Instructional designers become experience architects. Small businesses are moving from content creators to knowledge verifiers. QA extends into AI governance as a built-in quality and compliance function, rather than as a bottleneck. This is the part of workforce transformation that most AI learning strategies completely miss.
They have governance built in from the beginning. Mature organizations define clear boundaries around the extent to which AI can be generative and the extent to which it must remain deterministic. Audit trails and traceability ensure that innovation does not compromise trust. This is especially important in regulated industries.
And the measurement methods are also different. Rather than tracking content volume or completion rates, track skill progression, internal mobility, and performance improvements. They answer a question that matters to business leaders: Are employees becoming more competent in the things that drive business outcomes?
where to start
For organizations looking to enhance their AI learning strategy, the starting point isn’t new tools or new platforms. This is an honest diagnosis.
Three questions worth asking:
Is your content structured in a way that supports modular reuse and AI interaction, or is it locked into a format designed for one-time delivery? Is your learning workflow designed for continuous evolution, or do you operate around budget cycles and project timelines that make continuous improvement structurally difficult? Is governance built into how you use AI in your content supply chain, or is it applied after the fact, creating hesitance that hinders scaling?
Answering these questions honestly will give you a clearer roadmap than any technology assessment. For organizations requiring structured benchmarking, Harbinger’s CLEAR Content Audit Framework provides scored diagnostics across content quality, AI readiness, learner experience, and library rationalization.
lastly
The future of learning isn’t determined by how fast you can create content. It is defined by how effectively an organization can create a system that continuously and at scale develops the capabilities of its actual workforce in alignment with its business direction.
At Harbinger, we work at the intersection of digital publishing, workforce enablement, and talent transformation. What we consistently find is that organizations that are making the most of AI for learning have one thing in common. That means you’re investing in the foundation before you invest in the features.
AI is a powerful enabler of workforce transformation.
But only if your system is ready to receive it.
