AI training for your team
SolidarIA trains teams at NGOs, nonprofits and foundations to use artificial intelligence: hands-on, by role, on your own documents. We measure what changes in practice, instead of counting the people trained.
- Format
- Half-days or a full day, on site or remote
- Group
- 12 people at most, at least 60% practice
- Tools
- Copilot, ChatGPT and Claude; Gemini on request
Which training should we choose?
Getting started with a tool
- The question
- How do I use it well, every day, in my work?
- The tool
- Your organisation’s: Copilot, ChatGPT or Claude; Gemini on request
- Who it is for
- All staff
- Safety rules
- 45 minutes at the start, non-negotiable
Responsible AI by role
- The question
- Where to use it, where to do without, and how to protect people?
- The tool
- Any tool
- Who it is for
- By role: programmes, monitoring and evaluation, protection, communication, leadership
- Safety rules
- A full half-day: the common core
How does getting started work?
One day, or two half-days. The outline adapts to your tool: Word, Outlook and Teams in Microsoft 365, shared projects in ChatGPT or Claude, Docs and Gmail in Google Workspace.
Before: the settings
2 h
With the person who runs your IT: accounts, sharing, retention, connectors, memory, what to switch off. It ends with a one-page sheet.
The ground rules
45 min
What you never paste in, checking before you share, saying when AI was used if the rules ask for it. Your rules, or interim ones.
Getting started
1 h 30
Interface, choosing a model, files, search and sources, shared spaces, memory. Writing a prompt: context, task, format, example, criteria. Iterate, then check.
Everyday workshops
3 h
About ten cases: each person picks the ones that fit their job.
Library and personal plan
45 min
The team stores 15 to 30 prompts in the tool. Each person picks two tasks to change within 30 days.
30-day follow-up
1 h 30, remote
You bring your real cases and what got in the way; we measure what has changed.
Everyday cases
- Emails and notes: drafting, rewording, changing the tone, replying in English.
- Meetings: minutes from notes, or from the transcript if you allow it.
- Reading fast: summarising a 60-page report, comparing two versions, pulling out the decisions.
- Proposal writing: a first draft of a section, fitting the donor’s template, checking against the logframe.
- Narrative reports from monitoring data, checking every figure.
- Spreadsheets: analysing an Excel file, writing a formula, making a chart.
- Translation into French, English or Arabic, with your glossary.
- Presentations: outline and slide content.
- Desk research with sources you can check, and its limits.
- Reusable assistants for repeated tasks, depending on the tool.
We train on work accounts. Without a work account, on fictional data only.
Which modules by role?
A common core, then one module per role.
Core: using AI without exposing anyone
All staff
Half a dayWhat a language model does, what it does well and badly, the mistakes it makes up. Three rules: never paste sensitive data into a tool that has not been approved, check before you share, say when AI was used if the rules ask for it.
Afterwards, they can
Spot a mistake, know what never to paste in, apply your rules.
Programmes and writing
Programme officers, coordination
Half a dayProposals, narrative reports, minutes; consistency with the logframe; the donor’s rules on AI.
Afterwards, they can
Produce a useful first draft, and check it.
Monitoring, evaluation and data
MEL teams
Half a dayStructuring data, coding interviews, anonymising; made-up figures and bias; documenting the use of AI in a methodology.
Afterwards, they can
Use AI without weakening the validity of an evaluation.
Protection, legal, case management
Lawyers, protection, social support
Half a dayWhat you never do there (victims’ files, children), acceptable uses (monitoring the news, blank templates, translating public texts), safeguarding and consent.
Afterwards, they can
Tell an acceptable use from a dangerous one.
Communication, advocacy, fundraising
Communication, fundraising
Half a dayTexts and translations; generated images and the debate on “synthetic poverty”; rules on dignity and transparency.
Afterwards, they can
Produce content without harming donors’ trust or people’s dignity.
Administration, HR, finance
Support functions
Half a dayDocument templates, summaries; care with staff data.
Afterwards, they can
Save time on repetitive tasks without exposing data.
Leadership and board
Leadership, board members
Two to three hoursWhere, why and who decides; risks; buying tools; obligations (data protection, the EU AI Act for organisations based in the EU); measuring the effects.
Afterwards, they can
Decide and steer, not just authorise.
Training internal champions
One or two people per office
Two days, then two monthly check-insRunning the core module themselves, in their own language, with the kit provided. An answer to field languages, and to the cost for an organisation working in several countries.
Afterwards, they can
Carry the training out into the field.
How do you train?
At least 60% practice, 12 people at most per group. Beyond that, it turns into a lecture.
On your documents, anonymised beforehand by you with a checklist we provide. Failing that, on realistic fictional documents.
On the tools you allow. If you allow none, on fictional data only.
Your teams’ objections become learning goals: jobs, privacy, the environment.
No vendor promoted. We compare, we don’t sell.
Remotely, a version built for weak connections: light materials, exercises that work offline.
How do you measure the effect?
We follow the four levels of the Kirkpatrick model, and say honestly how far a training can go.
1. Reaction
At the end of the session
Usefulness, clarity, what is missing.
Three questions
2. Learning
Before and after
Knowledge, and above all the ability to spot a mistake.
A ten-question test, and an AI-written text with five deliberate mistakes: a made-up figure, a non-existent source, a forgotten piece of personal data, a bias, a misreading
3. Behaviour
30 days later
Practices taken up, risky practices dropped, obstacles.
A short survey and a 30-minute group call
4. Results
From 3 to 12 months, within AI steering
Time saved, quality, incidents.
Indicators tracked over time
The ten-question test is the same as in the diagnostic: if you do both, you see your progress on a single scale.
For getting started, we add three timed tasks in the tool, scored on a grid (faithfulness, accuracy, format), before and after. At 30 days, time saved is shown for what it is: self-reported.
Time savings are never promised for a training on its own: a single day does not let anyone attribute them honestly.
What does the training deliver?
The materials, adapted to your tool and dated, and a one-page cheat sheet for each participant.
A “privacy settings for your tool” sheet.
The team’s prompt library, stored in the tool. It belongs to you.
An individual certificate: content, length, date, trainer. For an organisation based in the EU, it is the written record recommended under Article 4 of the EU AI Act.
A 2 to 3-page report: attendance, scores before and after (aggregated and anonymous), 30-day results, recommendations.
If you don’t have an AI policy yet: one page of interim ground rules, adopted during the session.
Which training format suits us?
Tool pathway
- What it includes
- The settings session, two half-days, the 30-day follow-up, the tests before and after.
- Group
- 12 people at most
- Length
- Two half-days, then a 30-day follow-up
Getting-started half-day
- What it includes
- The ground rules, getting started and three practical cases.
- Group
- 12 people at most
- Length
- Half a day
Cohort of several organisations
- What it includes
- Four 90-minute remote sessions on the same tool, then the 30-day follow-up.
- Group
- 8 to 15 people
- Length
- Four remote sessions
Settings session only
- What it includes
- Two hours with the person who runs your IT, and the settings sheet.
- Group
- Your IT team
- Length
- Two hours
Role modules take half a day, champion training two days. Licences are not included, and we earn nothing on them. Travel costs are extra. The solidarity rate applies.
Request a quoteAre you tied to a vendor?
No. Training on Copilot, ChatGPT or Claude does not undermine independence, on these conditions.
You choose your tool, or we help you choose it with a neutral grid.
No commission, no resold licence.
If SolidarIA ever joins a vendor’s partner programme, we will tell you.
Never a training on a consumer account with real data.
How does it fit with the rest?
The diagnostic tells you which tool and which subscription to choose.
The AI policy provides the ground rules for the first session.
The role modules go deeper, once the tool is mastered.
Twelve-month AI steering tracks use over time, with indicators.
SolidarIA - Académie offers free online courses on AI, in French and with no sign-up: everyone takes them at their own pace.
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 which training would help you. You can also request a quote right away.