Published 03 Feb 2025 · AI Centre of Excellence

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.

Lire en français

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.

Updates