Automate what should be automated. Escalate what requires human judgement.
Responsible AI is part of service quality and safeguarding, not a compliance appendix. A service that answers quickly but wrongly, or keeps a person in an automated loop when they need help, has failed them.
Our principles
People stay in charge
- Human oversight of automated journeys
- Human escalation with agreed response times
- Safeguarding procedures owned by partners
Trusted content
- Approved, partner-owned knowledge sources
- Testing and validation before launch
- Localisation for language and context
- Version governance of all content
- Transparency that people are using an automated service
Protected data
- Data minimisation: only what the service needs
- Informed consent
- Privacy by design
- Role-based access and secure integration
- Auditability of decisions and data access
Sensitive decisions stay with accountable people.
In high-sensitivity areas, automation supports information, navigation and data capture. Decisions stay with qualified people, and every journey is designed so a person can reach one.
- HealthClinical questions and treatment decisions
- SafeguardingAbuse, exploitation or safety concerns
- Financial servicesCredit, suitability, claims and disputes
- Humanitarian servicesProtection concerns and eligibility disputes
The system should never invent requirements, rules or advice.
- Answers are retrieved from approved, partner-owned knowledge sources, with deterministic workflows where accuracy matters most.
- Content is version-controlled, so every answer can be traced to the approved source in use at the time.
- Each journey has a defined escalation route, with response times agreed with the partner.
- Unanswered questions are logged and reviewed, so content gaps are closed rather than guessed at.
- People are told when they are using an automated service, and how to reach a person.
- We design to comply with applicable data-protection law in each jurisdiction where a programme operates, and with partner and funder data-governance requirements.
What stands between people and the service you need them to reach?
We work with governments, development partners and implementing organisations to identify that gap, design an accessible service journey and test what works.