AI in your firm,
without breaking your professional ethics.
Lawyers, notaries, accountants, healthcare: your profession requires professional secrecy and strict ethical rules. We design AI use cases that fit inside that framework, not next to it.
Colombani.ai is a consulting and training firm dedicated to regulated professions in France: lawyers, notaries, accountants, healthcare professionals. Our approach addresses professional secrecy and your order's ethics (CNB, CSN, OEC, CNOM) first, then model selection by data sensitivity (EU Claude, local Mistral, on-premise AI). Qualiopi-certified training (NDA 11757549975) eligible for FIF-PL and OPCO EP funding, instructor holds Claude Certified Architect (Foundations) issued by Anthropic.
Why a dedicated approach
Professional secrecy is not negotiable
Sending a client file to ChatGPT, Claude or Mistral cloud without precaution can breach professional secrecy. Model choice, anonymization, local hosting: we handle this first, not last.
Your professional order already regulates your activity
CNB for lawyers, OEC for accountants, CNOM for physicians. Each order publishes or is preparing its position on AI. Our training covers the up-to-date stance of your order, not a generic framework.
AI Act + GDPR + sector-specific rules
You stack general GDPR, AI Act (which classifies certain healthcare or justice uses as high-risk), and sector obligations (HDS, AML, notarial forms). We identify what applies to YOU and what doesn't.
A page dedicated to your profession
Each profession has its own use cases, constraints and suitable tools. Pick yours.
Law firms
Case law analysis, drafting briefs, legal monitoring. With strict respect of professional secrecy.
Notarial offices
Deed consistency checks, regulatory research, automated file sorting.
Accounting firms
Account analysis, summary notes drafting, automated tax monitoring, AML support.
Healthcare professions
Patient records, file anonymization, therapeutic monitoring. With HDS hosting and local AI when required.
A use case in mind?
Describe your situation and the type of data involved. Straight answer within 48 hours on what is feasible and what is not.