Case Study: GitLab achieves secure role-based AI onboarding with Kipwise

GitLab improves AI onboarding in a 5-persona pilot with Kipwise

GitLab faced the challenge of personalizing AI onboarding assistance for new hires by role, team, and location without accidentally leaking internal knowledge from restricted documents. A prompt-only approach was insufficient as it could not enforce access controls, potentially exposing sensitive procedures to the wrong employees or contractors. They partnered with Kipwise to implement a secure and structured AI solution.

Kipwise implemented a solution that first resolves a user's trusted identity to build an authorization filter, ensuring the AI's retrieval layer only searches documents the employee is explicitly allowed to see. The solution then uses metadata like role and onboarding stage to rank these allowed sources for relevance. This approach provided personalized, secure guidance backed by citations to approved sources, refused unsafe queries, and routed gaps to content owners, enabling a successful pilot and secure rollout.

GitLab