Responsible AI diagnostic
The responsible AI diagnostic is a two-to-four-week AI readiness assessment for NGOs, nonprofits and foundations. We look at what your teams really do with artificial intelligence, where your data goes and what could go wrong. We measure your starting point, then measure it again six months later.
- Length
- Two to four weeks, depending on your size
- Method
- Anonymous survey, interviews, feedback workshop
- Languages
- Survey in French and English; Arabic or Creole as an option, at the cost of translation
What do you look at?
Actual uses
Who uses what, for which tasks, with which account, including free personal accounts.
What comes out
A map of uses by role.
Data and protected people
Which sensitive data goes into which tools, and where it ends up: beneficiaries, victims, children, health, location.
What comes out
A map of data flows, and the uses to stop right away.
Governance
Who decides, which rules exist, how you buy a tool, what your safeguarding policy says.
What comes out
A score on six governance elements.
Skills
What your teams actually know, not just what they say they know.
What comes out
The level by role and the training needs.
Risks
What could go wrong, for whom, and how badly. Harm to the people you serve comes before everything else.
What comes out
A risk register: for each risk, a measure, a person in charge and a deadline.
Opportunities
Where AI would save you time without exposing anyone, and where you are better off without it.
What comes out
Three to five priority uses, and the ones to set aside.
How do you measure the baseline?
Eight indicators, measured during the diagnostic. Six months later, we run the same survey again and talk it through for an hour: you see what has moved.
Share of staff using AI every week, including on a personal account
Anonymous survey
Number of uses that touch sensitive data
Inventory and interviews
Share of uses covered by a written rule
Uses checked against your documents
Governance score out of six: policy, named lead, tool register, purchasing procedure, contract clause, incident reporting channel
Document review
AI literacy, self-reported and tested, by role
Self-assessment and a ten-question test
Weekly time spent on two or three tasks: donor reports, minutes, translations
Self-reported, ideally with one day of time logging
Your teams’ trust in AI and their concerns
Scale in the survey
Known incidents: leak, published error, complaint
Interviews
How does the diagnostic work?
Scoping
One hour with your leadership. You name an internal contact, we sign a confidentiality agreement and list the useful documents.
Listening
An anonymous survey open to all staff for ten days, one-hour interviews, a review of your documents and an inventory of your tools.
Analysis
We observe two real tasks, build the risk register and calculate the indicators.
Feedback
A two-hour workshop with leadership and the team. We set priorities together and write down the decisions. The report follows.
Six months later
We run the same survey again, then talk it through for an hour.
What we ask of you
- An internal contact, about four hours over the whole period.
- Two hours of your leadership’s time.
- Sharing the survey with your teams.
- The list of your software subscriptions and your existing policies.
- Before the survey, a written commitment from your leadership: no sanction for the uses people report. Without it, the answers would not be honest.
What does the diagnostic deliver?
A 15 to 20-page report, and a 2-page summary for your board or your donors.
A map of your uses and your data flows.
The risk register, with a person in charge and a deadline for each risk.
A “stop on Monday” list.
A twelve-month roadmap in three stages: right away, at three months, at twelve months. Each action says what it costs and what it risks.
The indicator table and how it is calculated, so you can measure again.
The survey, to reuse. Everything produced belongs to you.
Which diagnostic format suits us?
Essential
- For whom
- Up to about thirty people, on a single site.
- What it includes
- Survey, six interviews, two observed tasks, one feedback workshop.
- Length
- Two to three weeks
Full
- For whom
- Several sites or several countries.
- What it includes
- Up to twelve interviews, two languages, two feedback workshops: leadership, then teams.
- Length
- Three to four weeks
Donor or network
- For whom
- A donor funding the diagnostic of four to six partners.
- What it includes
- One Essential diagnostic per organisation, plus a consolidated report for the donor.
- Length
- Six to eight weeks
Solidarity rate on the Essential and Full formats: 25% or 50% off, depending on your annual budget.
Request a quoteWhat happens to our data?
The diagnostic follows the rules of an evaluation, such as the ethical guidelines of the United Nations Evaluation Group (UNEG).
We never access your beneficiaries’ data: only inventories and metadata.
The survey is anonymous.
Every interview starts with explicit consent.
The diagnostic data is deleted after a period set out at the start.
The data grid follows the Inter-Agency Standing Committee (IASC) operational guidance on data responsibility, 2023 edition. For organisations based in the European Union, the report also keeps the written record of AI literacy required by Article 4 of the EU AI Act since 2 February 2025.
What is the diagnostic not?
Not a legal audit
Compliance questions are flagged as points of attention, to be checked by a lawyer or your data protection officer.
Not an IT security test
We do not look for flaws in your systems.
Not a tool sale
We can compare solutions, with no vendor commission.
Not a technical evaluation
We do not measure how well an AI model performs.
How does it fit with the rest?
The diagnostic says which tool and subscription to choose: training starts from there.
It lightens the preparation of the AI policy, because your real uses are already known.
It sets the baseline that twelve-month AI steering measures again.
Who leads this work?
Arnaud Dandoy, PhD in criminology (University of Kent) and law graduate (UCLouvain). More than ten years with public institutions, the United Nations and international NGOs, in Tunisia, Morocco and Haiti. Based in Tunis, working remotely and on site.
Read the full backgroundWhere do we start?
The free hour is enough to see whether the diagnostic would help you. You can also request a quote right away.