AI Training package
Corporate AI Training: against Shadow AI
A structured programme to understand what your employees are really doing with AI, train them on enterprise-grade models and tools, and turn experimentation into high-value vertical agents. Covers the AI Act art. 4 training obligation.
The problem: Shadow AI is already inside your company
Your employees are already using consumer AI tools (ChatGPT Free, Claude.ai, personal Microsoft Copilot), often from personal accounts, often pasting contracts, payslips, client emails and strategic plans into prompts. This is Shadow AI: untracked, ungoverned use of AI tools inside the company. It's not bad intent: for many employees it's simply the only way to get their work done in reasonable time. But it happens outside the control of IT, the DPO and management.
Banning ChatGPT almost never works: people keep using it from their personal phone. What you need is a governed alternative and a programme that brings employees inside the perimeter, not outside.
What you're risking if you don't act
Three open fronts, all already measurable today.
GDPR compliance and privacy
Personal data and proprietary information end up in consumer services with standard TOS that don't satisfy GDPR and without adequate DPAs. Often on extra-EU infrastructure, with no tracking of transfers, retention or processing.
AI Act, in particular art. 4
The AI Act is in force. Article 4 imposes AI literacy obligations on employees who use AI tools at work. Without a documented and traceable training programme it's impossible to demonstrate compliance, classify the systems in use or document the risks.
Quality and reproducibility
Untracked AI outputs end up in corporate documents, operational decisions, customer communications. Nobody knows who generated them, with which prompt, with which model. When a problem emerges (an error, a bias, a hallucination) debugging is impossible.
The four-step programme
No generic courses. 8-12 weeks for the full programme, with first deliverables already after 4 weeks.
Step 1
On-site training
An AIDAPT trainer on site dedicated to your teams. Content calibrated on seniority and corporate functions. Traceable training documentation, useful to cover the AI Act art. 4 obligation.
Step 2
Experimentation software
A secure multi-model sandbox (OpenAI, Anthropic, Google, Mistral) where your employees experiment on company data without leaving the compliance perimeter. The governed alternative that makes consumer tools unnecessary.
Step 3
Conversation analytics
We analyse, in aggregated and anonymised form, what staff are trying to do with AI: which processes, which blockers, which recurring patterns. This is where it becomes clear where AI really makes a difference in your company, and where it's just noise.
Step 4
Identifying high-value vertical agents
Shortlist of vertical AI agents to build, with priority, impact estimate and implementation roadmap. From here we continue with custom development or the Caity platform: only if it makes sense, only where it makes sense.
Frequently asked questions
SMEs and mid-market companies between 20 and 500 employees, with at least one knowledge-intensive department (legal, HR, operations, finance, customer care). Sounds like you?
No, in itself it isn't illegal. But the way it often manifests (personal data in consumer services without an adequate contract) can violate GDPR. And in many cases it amounts to a breach of the AI Act in terms of governance and traceability.
Usually no. People keep using it from their personal phone or private account. A 'do not use' policy without a governed alternative is almost always ineffective. Our approach is the opposite: bring them inside, don't leave them outside.
The full programme runs 8-12 weeks. The first concrete deliverables arrive already after 4 weeks.
Yes. Dedicated deployment or isolated tenants, conversation analysis in aggregated and anonymised form. No data leaves the agreed compliance perimeter.
OpenAI (GPT-4.x), Anthropic (Claude 4.x), Google (Gemini), Mistral, and open-source on-premise models where needed.
Yes. The programme is designed to produce traceable training documentation, useful in case of audit or verification.
Fixed onboarding fee + licences for the experimentation software. No per-user fee. Quote shared after the scoping call.
Shortlist of vertical agents → custom development or Caity platform, only if needed. No obligation to continue: the programme has value in itself, independently of any follow-on projects.
Ready to clean up Shadow AI?
Talk to our team: together we'll decide if it makes sense to start the programme, no commitment.
