Generative AI Pilot Results: Summary Report
Findings from a six-month generative AI pilot across three federal departments, migrated from AICoE's pilot reports archive.
This report summarizes outcomes from a six-month generative AI pilot conducted with three federal departments, migrated from the AI Centre of Excellence's pilot reports archive.
Overview
The pilot tested generative AI assistance for drafting internal correspondence, summarizing meeting notes, and triaging service requests.
Findings
- Time savings averaged 22% on drafting tasks.
- Human review remained mandatory for all external-facing content.
- Bilingual output quality required additional prompt tuning for French.
Recommendations
Departments considering similar pilots should budget time for prompt engineering in both official languages and establish a review workflow before scaling beyond a pilot cohort.