Algorithmic Risks in Corporate Learning

Williamson (2016) raises key concerns about how big data is reshaping education in his blog. While his focus is largely on schools and universities, the questions he poses has a growing relevance in workplace learning environments as well. Especially in corporate settings where educational technologies are deployed with high expectations for efficiency.

In my role as a learning consultant in the financial services sector, I’ve seen this unfold firsthand. We have been testing out the integration of Microsoft Copilot, an AI assistant embedded in tools like Teams and Word that pulls information from internal company documentation. While it’s framed as a way to support productivity and self-directed learning, I have noticed that it often returns outdated, incomplete, or unverified information.

Williamson’s question “How are machine learning systems used in education being ‘trained’ and ‘taught’?” has direct relevance here (2016). From my understanding as a user so far, Copilot has been trained on available internal files and uses a large grouping of data from Sharepoint files to internal working dockets to pull information from. Although this information is internal and there are limited risks to external exposure, there doesn’t seem to be a formal vetting process, instructional oversight, or accountability for how it interprets that content. In one instance, I’ve seen a colleague make a recommendation based on Copilot’s summary, only to discover the source material was three years old and deprecated. That moment revealed how the illusion of authority in algorithmic tools can have real consequences.

More concerning is the impact on the learners themselves. This passive engagement can undermine not only learning depth but also critical thinking skills, especially when learners are unaware of the algorithmic limitations at play.

If we continue to adopt AI-based tools in corporate learning without structured oversight, data literacy, or ethical design practices, we risk embedding long-term inequities and misinformation into daily workflows. Williamson’s call for data accountability should not stop in the classroom. It also belongs within workplace settings and L&D strategy discussions as well.

Williamson, B. (2016). Critical questions for big data in Education.  https://codeactsineducation.wordpress.com/2016/06/02/critical-questions-for-big-data-in-education/