Principles first, tools second
Durable controls that do not depend on any particular vendor or model, with an updatable tool annex.
Case study 05IPPF · Global responsible AI
International Planned Parenthood Federation
Federation-wide guidance on the use of AI for research.
A governance-first, participatory process that turns existing good practice and IPPF's own policies into usable decisions for researchers, MEL teams and Member Associations across every region.
- Role
- Team Lead, Quantum AI R&D
- Scope
- Global MA consultation
- Regions
- All five non-EN regions
- Delivery
- Trinidad & Tobago + Cyprus
The challenge
IPPF needed more than an AI policy. It needed operational guidance for a federated, values-led organization where research spans the full life cycle — monitoring, evaluation, learning and everyday evidence work.
The guidance had to enable proportionate use while protecting service users, participants, staff and organizational trust across many jurisdictions.
The approach
How the work is designed.
Risk follows data and task
Classification, purpose, affected people, reversibility and required assurance determine the control level.
Honest picture of use
Non-punitive, anonymised mapping that separates sanctioned use, shadow use, unmet needs and incidents.
Federation-wide relevance
Consultations across all five non-English regions with purposeful inclusion and documented contextual variation.
Human-in-the-loop, defined
An accountability mechanism with named reviewers and approval gates — not a generic disclaimer.
A living product
Named ownership, versioning, a six-month review cycle, adoption indicators and reusable adaptation tools.
How the proposal maps to IPPF's needs
| IPPF need | Response | Deliverable |
|---|---|---|
| Safe, practical guidance | Technology-agnostic rules | Decision trees, worked examples, cheat sheets |
| Protection of people and data | Service-user data boundary | Classification gates, human review, DPO sign-off |
| Federation-wide relevance | Regional consultation | Documented contextual variation |
| Living product | Governance ownership | Versioning and six-month review cycle |
Governance built in
- Personal records and interview material stay out of scope unless a named use case completes DPIA, ethics, DPO and MA approvals.
- Good practice is attributed only with consent; risks, incidents and shadow use are never attributed.
- Legal interpretation remains with IPPF Legal and the DPO.
Where it stands
- One evidence chain connecting mapping and consultation to principles, controls, examples and adoption tools.
- Reusable adaptation tools for Member Associations.
- A governance product designed to evolve with the technology.
"Responsible AI has to be usable, or it will be ignored."Book a strategy call
Next case study
Quantum AI R&D