The workplace seeks solutions outside the rules
AI is no longer a hype. It has become a daily reality in the workplace. But while the technology is developing rapidly, the formal AI policy of many companies is lagging behind. According to the recent STEM Workforce report from SThree, 72% of Dutch tech professionals use AI tools that are not officially approved by their organization. Think of ChatGPT, Gemini, or Copilot: public platforms that find their way into operational processes without internal validation.
Why does this happen? Because it works. Employees are overwhelmingly choosing these tools due to the speed, ease of use, and functionality that approved alternatives do not (yet) offer. But the downside is significant, especially in terms of security, compliance, and control.
The biggest pitfall: apparent efficiency
What starts as a productivity gain often ends in dependency. Nearly a quarter of respondents in the survey indicate that they cannot complete their work without unapproved AI solutions. If the AI is removed, work comes to a halt. The tool thus becomes mission-critical, without the organization having consciously allowed or protected it.
The real pitfall lies in the invisibility. IT departments have no visibility into what is being used, what data is being shared, or how secure the storage is. This puts companies at unconscious risk of data breaches, reputational damage, and violations of laws and regulations such as the GDPR and soon the EU AI Act.
The risks: from privacy to policy damage
The risks are not hypothetical. 81% of professionals acknowledge that unauthorized AI use can jeopardize privacy or security. Yet the tools remain popular. Why? Because the official alternatives are often too slow, too limited, or poorly integrated into the workflow.
This creates a paradox. Companies invest in secure infrastructures, but their own staff bypass these because the work simply cannot be done otherwise. Without proactive policy, organizations unconsciously let their data security depend on tools they have not chosen, tested, or validated themselves.
How can companies bridge this AI gap?
The solution lies not in banning but in guiding. Companies must actively invest in approved AI tools that are indeed usable and powerful. This requires collaboration between IT, security, and business units. Additionally, there must be clear communication about risks, governance, and alternatives.
The upcoming EU AI Act provides companies with a framework to establish responsible AI policy. This requires not only technical but also ethical and organizational choices. Those who do not provide employees with clear guidelines force them into shadow work. And thus into risk.
AI without oversight is an open backdoor
Generative AI is powerful. But without policy, it is also vulnerable. Companies that ignore AI at the front end lose control at the back end. The future lies in controlled freedom: AI solutions that are tailored to real work processes and meet the requirements of IT, security, and compliance.
Now is the time to move from reactivity to management. Unapproved AI tools are no longer an incident. They are a symptom of a structural lack of good policy.
