Skip to content

Ethics and independence

The rules I apply to every engagement, published so that clients and partners can hold me to them.

Independence

I do not independently evaluate, review or assure any program, policy or AI system I helped design, manage or advise on, and I do not evaluate a program whose reports I wrote. Where an advisory client requires independent assurance of work I contributed to, I will help commission a qualified third party and step back from that component. Evaluation engagements include a written declaration of any prior relationship with the program or organisation being evaluated.

Standards

Evaluation work follows the OECD DAC evaluation criteria and quality standards, the UNEG Norms and Standards, and the commissioning institution's own evaluation policy. Work on AI-enabled programs is referenced to published frameworks (ISO/IEC 42001, NIST AI RMF, the EU AI Act). Findings are reported as evidenced, including when they are unwelcome.

Confidentiality and data

Client material is handled on encrypted devices, shared only through channels the client approves, and retained no longer than the engagement and its contractual record-keeping require. Project references on this site are sanitised; no client logos are used and no institutional endorsement is implied by any reference to past work.

Conflicts

Current and recent engagements are disclosed to prospective clients where a reasonable person would consider them relevant. I do not hold standing advisory arrangements with organisations in direct competition with one another.

Bids and consortia

My CV is included in a proposal only with my written consent for that tender. Before agreeing, I check the program and country against my past and current work and decline where a conflict exists. Where a tender requires exclusivity, I commit to one bidder and say so to any other firm that asks.

AI use in delivery

I use AI tools in my own practice, under my own governance rules: client-identifying material is not entered into consumer AI tools, outputs are verified before use, and clients may set stricter conditions in the contract.

To discuss an assignment, get in touch.