The Operating Platform for Animal WelfareGet your free ARC Score
Get Your ARC Assessment

ARC Academy / AI

Responsible AI Readiness

Choose safe first use cases, protect sensitive information, and establish review rules before your team adopts AI.

Begin Lesson One All Courses

Course brief

Foundation55 minutes4 lessons

By the end, you can:

  • Screen potential AI use cases by risk and value.
  • Create a no-share data rule for staff and volunteers.
  • Define human review and escalation requirements.
0115 minChoose a bounded first use case

Objective

Start with work that is reversible, reviewable, and low risk.

The best first use case saves drafting or summarizing time without allowing AI to make eligibility, medical, legal, safety, or financial decisions.

Do the work

  1. List repetitive tasks that start from approved internal information.
  2. Remove tasks involving final medical, legal, safety, employment, or eligibility decisions.
  3. Score the remaining tasks for frequency, reviewability, reversibility, and sensitivity.
  4. Select one task with a clear human reviewer and a simple fallback process.

Leave with

A scored AI use-case shortlist with one approved pilot.

Reflection

Can a qualified person verify the entire output before anyone relies on it?

0215 minSet data boundaries

Objective

Make it unambiguous what information may and may not enter an AI tool.

A practical policy names prohibited data and approved tools in plain language. It does not rely on every user interpreting “sensitive” the same way.

Do the work

  1. Prohibit passwords, payment data, government identifiers, private medical details, and confidential personnel information.
  2. Decide how adopter, foster, volunteer, donor, and staff information must be anonymized.
  3. Name the approved tools and accounts; prohibit personal accounts for organizational work.
  4. Document how prompts and outputs may be retained, shared, or deleted.

Leave with

A one-page AI data-boundary policy for staff and volunteers.

Reflection

Could a new volunteer follow the rule without needing to guess?

0315 minDesign human review

Objective

Put accountable people between an AI draft and any consequential action.

Human review is a workflow, not a disclaimer. The reviewer needs authority, context, a checklist, and a clear escalation path.

Do the work

  1. Name the role qualified to review each use case.
  2. Create a review checklist for accuracy, missing context, tone, bias, privacy, and required disclosures.
  3. Define what the reviewer may approve, what must be revised, and what must be escalated.
  4. Keep an incident log for harmful, misleading, or privacy-sensitive outputs.

Leave with

A review checklist and escalation path for the pilot.

Reflection

Who is accountable if the AI output is wrong, and do they have enough context to catch it?

0410 minRun a two-week pilot

Objective

Measure whether the tool improves work without creating unacceptable risk.

Pilot with fictional or appropriately de-identified records first. Compare the AI-assisted workflow with the current process using the same quality standard.

Do the work

  1. Record baseline time and common errors for five representative tasks.
  2. Test with fictional or de-identified information before real operational use.
  3. Track drafting time, review time, corrections, escalations, and rejected outputs.
  4. At the end, choose expand, revise, pause, or stop—and document why.

Leave with

A pilot scorecard with time, quality, risk, and a documented decision.

Reflection

Did the pilot reduce total work, including review and correction time?

Ready for your organization

Turn the lesson into an implementation roadmap.

The free ARC Assessment turns your current systems, priorities, and constraints into a recommended starting plan.

Take the ARC Assessment