From “I Agree” to “I Understand”: Rethinking AI in Corporate Learning

In a world filled with many disagreements, the digital sphere is often a place of relative peace: There, we amicably agree and consent to almost everything… often without ever understanding what we’ve agreed to. One study has indicated that over 80% of users in the digital world click on “I agree” clickwrap prompts without ever reading or understanding them.

Never has there been such a disconnect between what we agree to and what we understand, and the application of AI is no exception to this. Up until 2024, the rapid expansion of AI in both corporate and consumer domains was largely unmonitored and unregulated. Even before AI and the EU AI Act, which aims to steer its application into safer waters, “I agree” buttons were becoming flashier, and we have taught ourselves to click them faster and faster, without actually reading and understanding which rights we are consenting to provide, and even worse… with whom we are agreeing.  

Enterprise L&D is already under significant pressure to incorporate some form of AI into its ecosystem, and it seems that the EU AI Act is not here to prohibit it, but to lead it in a more stable and secure direction. 

Giving conscious consent – EU AI Act implications

Agreeing to something means listening to it or reading it, understanding it, thinking about it, and giving our own stance on it. All these steps require a degree of literacy in the given topic. Therefore, AI literacy is one of the key points in the EU AI Act and an extremely important topic of Article 4. AI literacy does not only include the technical knowledge, which is often obscured from the end users who just want to access and share information. It also includes our conscious and informed use of AI systems, together with the interpretations, corrections, and oversight still required even in highly trained models. Since unregulated AI usage in corporations all around the EU has rapidly increased from 2021 up until 2024, building AI literacy in organisations has become a challenge in its own right. 

Since the speed of innovation in artificial intelligence is beginning to outpace basic cybersecurity practices, external regulations of AI usage cannot be expected to shoulder all the responsibility. According to Wiz researchers, 65% of companies on the Forbes AI 50 list had exposed verified secrets such as API keys, tokens and credentials. Because of vulnerabilities like these, nurturing responsible AI principles internally is a longer but more stable path for corporations to tread. Seeing the EU AI Act as an obligation creates even more friction between data security and innovation, but seeing it as a learning process can nurture the slowly-growing trust in AI-driven systems. 

AI and corporate learning should not “just agree”

Having a successful learning platform in any enterprise context is difficult without extensive knowledge sharing on core business processes, advantages and disadvantages of business models used, highly sensitive data on both processes and customers, etc. It is therefore easy to see how implementing third-party AI models without regulations such as the EU AI Act could be, and has already proved to be a serious security risk. When security risks outweigh the possible advantages which AI offers for enterprise L&D, sadly that is usually the tipping point when responsible AI principles for corporations are taken into consideration.

Of course, the EU AI Act already describes core principles of ethical AI usage which transfer well into the world of corporate learning. These include: 

• Local (on-device) processing to reduce risks of data leakage
• Minimisation and anonymisation of data transferred to third parties
• Traceability and logging for complete transparency over AI usage
• Keeping the decision-making on the side of the user rather than on the side of AI models

Next steps: When regulation meets learning strategy

Rather than restricting enterprises in their scaling, especially in the L&D and HR departments, AI regulation and HR strategy within the EU perimeter enables organisations to introduce Model Context Protocol (MCP) servers to help maintain data integrity, security, and privacy for all parties involved, but that is a topic for the next post in this blog series.

The only way enterprise learning and AI compliance in the EU should “agree” is through conscious consent and through building AI literacy first. Shifting responsibility from systems to human beings, from providers to users, from the collective to the individual is another area which requires special attention and development within organisations. To preserve user integrity and data security now that the era of abstract AI euphoria is over, the EU AI Act provides a sturdy basis upon which to build, especially in enterprise L&D.

In reality, the most reckless “I agree” in corporate learning is not clicked by accident, it is clicked deliberately, to move faster, to avoid friction, to keep the appearance of innovation alive. The EU AI Act is not a brake on progress; it is a stress test for leadership maturity. For enterprise L&D, the question is no longer whether AI can personalise, automate, or accelerate learning, but whether organisations are willing to stay accountable when it does. AI literacy is not a compliance checkbox; it is a power shift back to humans.

Those who treat regulation as an obstacle will outsource judgement to machines, while those who treat it as a framework will build trust, resilience, and long-term advantage. The trend of boundless AI enthusiasm without responsibility is coming to an end and in that sense, the future of AI in corporate learning does not begin with “I agree”. It begins with “I understand”. 

Talk to an expert