It may not be an exaggeration to say that artificial intelligence has brought a change to our everyday lives comparable to the emergence of the internet and social media. Naturally, the use of AI tools has also spread rapidly in the workplace: “Summarize this report briefly.” “Write an attention-grabbing email about the company sports day next Friday.” “Find the rules for requesting time off in this document.” There is no doubt that AI can save office employees a great deal of time and energy, but by now, the drawbacks have also become apparent. One of the greatest risks is that the boundaries between reliable, official company information and summaries or answers generated by AI, potentially through several steps, can become blurred.
Shadow AI refers to employees using external AI tools and applications that have not been approved by the company. Put simply, this happens when employees use these tools to access official company information, get answers to questions related to internal processes, or complete work-related tasks.
The motivation, of course, is usually efficiency: employees simply want to get things done faster. As a result, they may turn to a public chatbot and, for example, ask AI to explain a company policy. It is often unclear what the answer is based on, what company-related information has been copied into the AI tool as a source, or how reliable the result actually is. If the information obtained this way then starts circulating among colleagues, it becomes very difficult to trace where it originally came from and how trustworthy it is.
The other - more serious - issue is the data protection risk, for example when an employee wants to summarize a 20-page contract and therefore copies it into an AI application. If someone uploads company documents to external tools, data security can be seriously compromised. From employee and customer data to unpublished decisions, contractual information, or financial details, a wide range of sensitive information may be exposed.
To address the issues above, many companies are starting to develop internal chatbots or AI assistants. This may seem like an obvious solution, as employees are less likely to turn to external tools with uncertain reliability if they have access to their own internal AI. Especially one that can access HR documents, policies, rules and frequently asked questions, allowing employees to ask targeted questions. However, a chatbot is only as reliable as the information it works from.
If, for example, several documents exist on the same topic, does the AI know which one is current? What happens if a policy has not yet been updated, but teams have already been informed about the change?
Artificial intelligence can hide these problems by always providing an answer - employees may not necessarily question the reliability of AI even if they remember something different from an email or a conversation. This is why companies should not rely solely on artificial intelligence. Human oversight is necessary to ensure that AI generates answers based on current, approved and official company sources.
Naturally, a news update about a community event does not require the same level of human review as an occupational safety or HR policy. The former is low risk, while interpreting a policy related to labor law, data protection or occupational safety belongs to an entirely different category.
AI alone cannot solve the challenges of company information management: reliable answers require an organized, up-to-date and accessible information environment. In fact, the more automated answers appear within an organization, the more valuable it becomes for employees to know exactly where they can find reliable workplace information.
Blue Colibri makes approved company information available through a single central platform, while allowing every piece of content to be targeted to the relevant employees. Access permissions can determine who can see which information, while a central knowledge base can reduce the likelihood of outdated or duplicate information circulating within the organization.
On the Blue Colibri platform, employees can easily search for all information and policies officially considered valid by the organization. They can return to this source whenever they become confused by the different answers they receive from AI tools or colleagues.
