The rapid integration of Generative Artificial Intelligence (AI) into the corporate training landscape has fundamentally altered the economics of Learning and Development (L&D). As organizations rush to adopt large language models (LLMs) and automated authoring tools, the industry is witnessing a shift toward unprecedented production speeds and significantly lower costs. However, a growing body of evidence and professional critiques suggest that this efficiency comes at a steep price: the erosion of pedagogical depth and the neglect of actual skill development. While AI has successfully made eLearning cheaper, it has not inherently made it better, leading to a proliferation of "shallow" content that prioritizes completion metrics over cognitive capability.
The Efficiency Paradox in Corporate Training
The current state of eLearning is characterized by a paradox where the ease of content generation is inversely proportional to the quality of the learning experience. Instructional designer Beth Pennington, M.S., recently highlighted this issue through a personal account of a Six Sigma White Belt course delivered via an AI-powered platform. Expecting a structured, application-heavy curriculum, Pennington instead encountered a series of dull, AI-generated slides that lacked depth, real-world context, and instructional value. This experience is becoming increasingly common as organizations prioritize the "volume" of training over the "value" of instruction.
Generative AI didn’t invent poor instructional design, but it has acted as a massive force multiplier for it. In the past, creating a mediocre course required weeks of human effort, which often acted as a natural bottleneck. Today, that same mediocrity can be produced in minutes. The problem is not the technology itself, but the organizational incentives that favor rapid output. When the goal of an L&D department is simply to "check a box" or fulfill a compliance requirement, AI becomes the perfect tool to manufacture the appearance of education without the substance of learning.
A Chronology of AI Integration in Learning and Development
To understand the current friction between AI and quality, it is necessary to examine the timeline of how automation has entered the educational sphere:
- The Pre-AI Era (2000s–2010s): Focus was on the Learning Management System (LMS) and SCORM standards. Content was human-authored, often linear, and expensive to produce.
- The Rise of Adaptive Learning (2015–2020): Early AI applications focused on "branching logic" and basic algorithms to recommend content based on past performance.
- The Generative Explosion (2022–Present): The release of ChatGPT and subsequent LLMs allowed for the automated generation of scripts, images, and even entire course structures.
- The "Cheap Content" Phase (Current): We are currently in a period where the barrier to entry for content creation has vanished. This has led to an influx of "AI-first" platforms that promise course creation in "seconds," often at the expense of instructional integrity.
This timeline shows a clear shift from technology as a support for human designers to technology as a replacement for the design process itself.
Supporting Data: The Gap Between Production and Impact
The metrics used to judge the success of corporate training are often disconnected from actual business outcomes. A 2025 survey conducted by the Association for Talent Development (ATD), involving 277 talent development professionals, revealed a startling disconnect in how organizations measure success. According to the report, 93% of organizations diligently collect data on participant satisfaction—essentially asking if the employees liked the course. However, only 30% of those same organizations reported being proficient at using learning data to inform actual business decisions.
This data suggests that most organizations are measuring "participation" rather than "proficiency." AI-generated content is highly optimized for participation; it is easy to consume, quick to finish, and simple to track. However, these metrics are "vanity metrics" that do not correlate with a workforce’s ability to perform complex tasks or solve novel problems.
Further research supports the idea that while AI saves time, it does not necessarily improve the end product’s efficacy. A randomized controlled trial by the Education Endowment Foundation found that teachers using ChatGPT significantly reduced their workload in lesson preparation. While this is a win for efficiency, a separate 2025 experimental study by Netland and colleagues involving 447 participants found no statistically significant difference in knowledge assessment scores between those who watched AI-generated instructional videos and those who watched human-produced ones. Interestingly, participants rated the human-produced videos more favorably, suggesting a "human touch" deficit that AI has yet to overcome.
The Science of Effective Instruction vs. AI Output
The fundamental issue with AI-generated eLearning is that it often confuses informing with teaching. In the world of instructional design, providing information is only the first step. True learning requires a series of scientifically backed interventions that AI often struggles to replicate without heavy human oversight.
The Importance of Practice and Feedback
Research by Dunlosky and colleagues (2013) identified "practice testing" and "distributed practice" as the most effective techniques for long-term retention. Many AI-generated courses rely on simple multiple-choice questions that test rote memorization rather than the application of knowledge. For a course to be effective, it must provide learners with opportunities to make mistakes in a safe environment and receive nuanced feedback.
The Feedback Loop
Hattie and Timperley (2007) emphasized that feedback is one of the most powerful influences on learning. However, for feedback to work, it must be specific to the learner’s unique misunderstanding. AI systems often provide generic "Correct/Incorrect" responses or repeat the same information already presented in the slide, failing to bridge the gap in the learner’s comprehension.
Transfer of Training
The ultimate goal of any corporate training is the "transfer of training"—the ability of the employee to apply what they learned to their actual job. A meta-analysis by Blume and colleagues found that this transfer is heavily dependent on the learner’s motivation and a supportive work environment. AI-generated "slideware" rarely addresses the motivational or environmental aspects of learning, focusing instead on the delivery of raw data.
Official Responses and Industry Reactions
The L&D community is currently divided on the path forward. Many "AI evangelists" argue that the speed and cost-saving benefits of AI allow for "just-in-time" training that was previously impossible. They point to the ability of AI to translate content into dozens of languages instantly, making global training more accessible.
Conversely, veteran instructional designers argue that the "democratization" of content creation is leading to a "race to the bottom." The reaction from professional bodies like the ATD and various eLearning industry leaders suggests a growing call for "AI Literacy" among designers. The consensus is shifting toward a "Human-in-the-Loop" (HITL) model, where AI handles the administrative and drafting tasks, but humans remain responsible for the pedagogical strategy, emotional resonance, and final quality assurance.
Industry experts warn that if the market continues to reward "cheap and fast," the long-term impact will be a "capability gap" where employees hold many certifications but possess few actual skills. This has led some organizations to move away from "off-the-shelf" AI content in favor of bespoke, human-curated experiences for high-stakes roles.
Broader Impact and Implications for the Future
The implications of the "Cheap but Not Good" trend extend beyond the HR department. In industries where safety and compliance are paramount—such as healthcare, aviation, or manufacturing—the failure of AI-generated training can have life-or-death consequences. If a safety course is generated by an AI that misses a subtle but critical nuance of a specific workplace hazard, the "efficiency" of that course becomes a liability.
Furthermore, there is a psychological impact on the workforce. When employees are repeatedly subjected to low-quality, AI-generated training, they begin to view all corporate learning as a waste of time. This leads to "disengagement by design," where learners find ways to bypass the content just to get the certificate, further devaluing the organization’s investment in talent development.
To rectify this, organizations must shift their focus from output to impact. The future of eLearning should not be measured by the number of modules produced or the speed of their delivery, but by the measurable improvement in employee performance. AI can be a powerful ally in this journey—helping to create realistic simulations, providing 24/7 tutoring, and analyzing data to find skill gaps—but only if it is guided by sound instructional principles.
In conclusion, AI has undeniably succeeded in making eLearning cheap. It has democratized the ability to create content and removed the technical barriers to entry. However, the "quality" of a learning experience is not a byproduct of the technology used to create it; it is a result of intentional design focused on human capability. As the novelty of Generative AI wears off, the organizations that thrive will be those that realize a cheap course is only a bargain if the employees actually learn something. The real value of AI in L&D lies not in its ability to replace the designer, but in its potential to augment the designer’s ability to create meaningful, impactful, and high-quality learning experiences.
