Your employees already use AI assistants. The question is whether they use approved ones, safely and well.
Shadow AI is the default
BCG's AI at Work 2025 survey found that 54% of respondents would use AI tools even if not authorized. Microsoft and LinkedIn found in 2024 that 78% of AI users bring their own AI tools to work. Banning AI rarely stops it; it pushes it out of sight.
What an approved assistant program includes
- The tool: an enterprise AI assistant with data protection (no training on your data, access controls, logs), ideally inside tools people already use.
- The policy: approved tools, data rules, disclosure, who to ask (see the AI policy kit).
- Training: role-specific, hands-on, with time to practise.
- Champions: a network of employees who share tips and prompts.
- Support: a channel for questions and requests for new tools.
- Measurement: active use, use cases, time saved, confidence and concerns.
HR's role
IT and security choose and secure the tool. HR owns the parts that decide whether it works: policy, training, manager behavior, change communication and the link to jobs and skills.
Common pitfalls
- Licenses for everyone, training for no one — usage drops after a month.
- A policy written in legal language nobody reads.
- No answer to "what does this mean for my job?" — fear kills adoption.
- Measuring logins instead of useful work.