The University of Manitoba Libraries (UML) embarked on an evaluation and implementation journey with Primo Research Assistant. This presentation shares the process of assessing the tool's usability, accuracy, and alignment with user needs, along with its subsequent rollout. A multidisciplinary working group conducted hands-on testing, posing diverse research questions to evaluate its strengths and limitations. Key findings highlighted its utility as a starting point for research while identifying challenges such as limited local relevance and overly confident responses. Implementation efforts focused on clear positioning of the tool as a supplementary resource, robust training for library staff, and transparent communication of its capabilities and constraints to users. Lessons learned include the importance of managing expectations around AI tools, emphasizing academic integrity, and integrating traditional library resources as complementary supports. This case study offers insights for institutions adopting AI-driven research tools.