EducationNot for Profit

Navigating AI Safety & Governance: Why We Tested Our Internal AI Policy Against Real-World Workflows

Two people looking at laptop screens
An AI policy is useless if it sits in a shared drive without being tested against real-world workflows.

As we put the final touches on the latest update to PN Digital’s internal AI guidelines, our main focus was ensuring policy met practice. Rather than dropping a static document into a folder, we hosted a team Lunch & Learn to put our frameworks to the test. We ran through live scenario testing, answered technical questions, and gathered honest feedback from our team on how these tools affect their day-to-day client work, data privacy, and IP protection.

Through this process, we refined three core operational guardrails:

  • Data Safeguards: Establishing strict protocols to protect client confidentiality and sensitive information.
  • Human Oversight: Ensuring every output undergoes thorough human review, quality control, and validation.
  • Responsible Adoption: Maintaining high standards around ethical tool usage, accuracy, and brand alignment.

We incorporated that direct feedback into our finalised documentation. To close the loop, we shared these updated guidelines with our key partners and collaborators, ensuring complete alignment on data privacy and operational transparency across every shared project we deliver.

Governance isn’t a one-time document – it’s an ongoing, active habit.