The average learning investment per employee in 2016 was $1,273, amounting to 34.1 hours, according to ATD, and companies worldwide will spend more than $365 billion on training in 2018, according to Training Industry, Inc. However, in 2014, Gartner (then CEB Global) reported that learning that is not applied on the job (i.e., “scrap learning”) comprises an estimated 45 percent of the training delivered in average companies. On a positive note, organizations that made an effort to measure scrap learning and focus on continual improvement in that area saw scrap learning decrease from 45 percent to 33 percent. While still a high number, that 12-percent reduction represents a savings of over $1.5M for a 10,000-employee company.

In a 2018 CGS survey, 62.2 percent of L&D leaders reported that their programs were either good or excellent. In contrast, over 60 percent of millennials surveyed indicated they didn’t receive or benefit from company training, and fewer than 40 percent of baby boomers felt that their management team was invested in their training and development.

While many L&D teams are focused on implementing solutions that curate additional content for learners, these statistics suggest that the problem isn’t more content – it’s fixing the content that’s already there. With such a large gap between training delivered and training applied on the job, what is the remedy?

Strategic applications of “new generation” artificial intelligence (AI) for learning can help reduce scrap learning in several areas, including AI-driven content optimization, learner behavioral modeling, predictive learning analytics and autonomous personalization of e-learning content. Of these applications, the first step in winning the battle against scrap learning is content optimization.

AI-Driven Content Analysis

Leveraging AI to reduce scrap learning begins with making sure the AI understands what is being taught, what is being tested and how that content compares to the learning objectives. To do so, the AI must first ingest the existing course, assessments and objectives and then scientifically discern key topics and their context at a granular level. From this information, the AI can produce a heatmap that visualizes topic progression throughout the material as well as several different gap analysis views to compare coverage of each topic in the assessments and objectives. Noticeable misalignments then inform areas of the course or assessments that should be changed so they align with each other and with the overall training objectives.

In the example below, the AI engine ingested the existing course material and identified the topics it found throughout. We can draw many observations from this example, but three obvious areas of concern are topics 5, 10 and 7.


In the first case, the coverage of topic 5 in the assessment is underrepresented compared to the objectives and the course content. Unless this topic is tested elsewhere – perhaps in another course in the curriculum – it represents an area where the course developers should change the assessment. Topic 10 has the opposite problem, where the topic is lightly covered in the learning objectives and the course but is a significant portion of the assessment. Unless there is a reasonable business explanation for this discrepancy, the assessment should likely have fewer questions centered around that topic. Finally, topic 7 makes up a substantial portion of the learning objectives but is only lightly taught and moderately tested. Unless this topic is covered more thoroughly in other courses in the curriculum, it is an area where both content and testing should be modified.

Leveraging AI to optimize training content is a necessary first step in reducing scrap learning. It can help L&D teams:

  • Distinguish between relevant and extraneous material
  • Assess pre-test efficacy against the learning objectives
  • Reduce human bias in curriculum design
  • Ensure that training achieves learning objectives

When coupled with AI-driven learner behavioral modeling and autonomous personalization, AI-driven content optimization will reduce scrap learning and help employees to learn more effectively – resulting in increased overall return on learning (ROL) for the organization.