Employee workload and AI use

Sophie, a 38-year-old property administrator, describes her workload as "enormous," managing over 500 real estate listings, emergencies, and "300, 400 emails in arrears." She criticizes the tools provided by her employer as "archaic" and her superiors as "disorganized," noting that they do not supply the necessary tools to respond effectively.

Personal AI adoption

Sophie has subscribed at her own expense to Claude, an AI assistant created by Anthropic, to draft responses to property owners and tenants. She checks legal points to ensure accuracy but considers the tool an "incredible time saver." This use falls under "shadow AI," meaning the use of AI outside the company's defined framework.

Widespread shadow AI

The phenomenon is "massive" and affects all sectors, public and private, according to Gérôme Billois, a cybersecurity expert at Wavestone. He explains that organizations often move more slowly than employees in adopting AI tools, as those offered by organizations are "subject to regulatory and budgetary constraints" and may not be as efficient as those available outside for a few dozen euros.

Employee AI usage statistics

According to the Work Trend Index barometer by Microsoft and LinkedIn, in the office, 75% of employees worldwide were already using AI at work in 2024, and 78% used their own tools. Camille, a 34-year-old architect, uses AI to improve her texts, summarize files, or prepare analyses.

AI as a productivity tool

She rereads, checks, and refines the results but estimates that the work is already done "80-90%." She says her bosses did not want to get involved, so she used general public tools to "manage, optimize and decide on the time saved."

Company policies and AI

Cécile Dejoux, director of the Observatory for Managerial and HR Transformations at the Conservatoire national des arts et métiers, explains that the absence of tools or instructions encourages such uses. Sometimes, companies "forbid" AI, but employees use it "to go faster" and want "to manage, optimize and decide on the time saved."

Lag in AI policies

The framework can also be "under construction," whether it be the ethical charter, the tools or the use cases. In the meantime, employees develop parallel uses, "under the cloak": "the pace is not the same" between the development of rules and experiments. Once the benefits are experienced, "it's a bit addictive," she observes. The tools subsequently provided by the employer may struggle to compete.

Security risks of AI

These uses, however, raise security questions, warns Gérôme Billois, particularly the risks of data leaks related to the sharing of confidential documents. On the data transmitted to AI, "you have to think as if you were in the real world," says Schéhérazade Abboub, a lawyer specializing in data law. "You wouldn't give all the data related to the company's trade secrets, its know-how. You wouldn't give your personal data..." she adds.

Employee perspectives on AI risks

Camille downplays this problem, stating that confidentiality is already an issue with the digital tools used daily, from "emails to Teams." According to her, employers underestimate the role of AI in daily work and the fact that employees are "overloaded."

Recommendations for managing AI

For Cécile Dejoux, to prevent the use of AI in secret, managers must "map" the needs, provide appropriate tools, and define a usage framework within the company. A 41-year-old communications manager, who wishes to remain anonymous, admits that "nobody talks about it openly" in the office, but "everyone does it! Who doesn't use ChatGPT? It's futile to want to ban it."