A PwC Hungary survey shows that a substantial share of AI tools used in the workplace operate outside corporate control. Among employees who apply AI in their work, 67% regularly or occasionally rely on personal, free or otherwise non-corporate tools. Additionally, 42% of respondents say their organisation has no internal policy on AI or they are unaware of such rules.
This pattern—often referred to as shadow AI—means that while employees may gain time savings, companies risk losing oversight of business data and information, creating legal, financial and reputational exposure.
What tasks do employees use AI for?
Information retrieval is the most common application: 78% of AI users employ the technology for question answering and quick information searches. Other frequent uses include:
- translation: 47%
- image generation: 35%
- copywriting: 34%
- data analysis: 31%
When it comes to work tasks, information search leads as well: 73% mentioned it, and 35% said they would keep only this AI function if forced to choose in the next six months. More complex applications that are deeply integrated into workflows remain relatively rare.
Efficiency gains and time savings
Among digitally active internet users surveyed, 47% use some form of AI in everyday life, with significant generational differences: 56–57% among Generation Z and Millennials (Y), and 35% among baby boomers. Forty-two percent of AI users apply the technology for both personal and work purposes; this share rises to 52% among respondents with tertiary education.
Reported time savings are notable: more than 78% of employees who use AI at work experienced at least a 10% time gain on relevant tasks; 38% reported savings of 30% or more, and a further 40% saw improvements between 10–30%. Respondents said the freed-up time can be reallocated to more complex, higher-value activities.
Individual practices where organisational governance is lacking
Organised use of AI is uncommon: 70% of employees who use AI for work adapt the tool to the current task without established prompts, templates or standard procedures. Only 12% use recurring prompts or templates for repetitive tasks, 10% report AI being partially embedded in standard workflows, and merely 3% reported team-level, organised adoption.
When unsure whether content may be shared with an AI tool, employees take varied approaches: 32% anonymise data, 30% use fictional examples, 25% avoid using AI, and only 9% choose an approved corporate tool. Seeking managerial or expert advice is rare—just 3% consult a manager or qualified specialist—while 23% decide based on their own judgment.
Individual safeguards such as anonymisation reduce exposure of personal data but do not prevent business plans, financial figures or other confidential materials from ending up in systems outside company control.
Gaps in awareness, training and governance
The survey also reveals uncertainty in self-assessed AI knowledge: answers on a 1–10 scale showed wide dispersion, suggesting many users cannot accurately judge their competence. This complicates understanding which risks are mitigated by practices like anonymisation and which remain. PwC recommends pre-training assessments to design targeted, role-specific education.
Dr. Szűcs László of Réti, Várszagi és Társai Ügyvédi Iroda (PwC Legal) highlighted: “In many companies AI use is already commonplace while appropriate regulatory, educational and control environments are missing. Employees therefore often decide on their own what data, for what purpose and with which tools they share.”
Governance is about enabling safe use, not banning AI
The study’s authors stress that the solution is not to restrict AI use. Rather, organisations should create environments with approved tools, clear internal rules and ongoing, role-specific training to support safe adoption. Effective AI governance must cover the organisation’s full AI portfolio, including tools employees adopt independently, and align data handling, security and compliance requirements.
“The biggest risk of shadow AI is not necessarily one off data breaches or an erroneous AI output appearing in corporate processes, but the gradual process by which an employer loses control over the information, decisions and workflows generated inside the organisation,” the expert warned.
Dr. Szűcs also recalled that the European Union’s AI regulation (the AI Act) presumes organisations must know the AI systems they use and operate suitable governance and risk-management mechanisms, documenting AI use where required. If employees widely use tools unknown to employers, fulfilling those obligations becomes substantially more difficult—particularly in industries subject to strict data handling or compliance standards.
Conclusion
PwC Hungary’s survey indicates that AI has become an important productivity tool at work, but the widespread use of non-corporate solutions and the lack of internal policies create significant risks. For companies, the recommended path is not prohibition but establishing transparent, adaptive and practice-oriented governance to enable both innovation and compliance.
