Skip to content

Responsible AI in Practice

Responsible AI training for practical aid workflows

For humanitarian, development, social-protection, governance, donor, and public-sector professionals who need to use AI faster, safely, and with better judgement.

Two facilitators

Every live session

Live, not webinar

Instructor-led practice

20 max

Capped cohort size

6 x 90 min

Over three weeks

Not pre-recorded. Not a webinar. Capped at 20 participants.

Next open cohort

Starts Tuesday 2 June 2026

Dates
June cohort Open now

2, 4, 9, 11, 16, 18 June 2026

July cohort Next intake

7, 9, 14, 16, 21, 23 July 2026

August cohort Coming soon

Dates to be confirmed

Schedule
Tuesdays and Thursdays, 16:00 East Africa Time
Sessions are spaced so you apply what you learned on real work between sessions. Not a boot camp. A three-week practice cycle with 20-30 minutes of practical between-session tasks.
Investment
EUR 350 per person EUR 280 for staff from national NGOs, local organisations, and aid workers between roles
Group size
Up to 20 participants

Live Intro Labs

1 June 2026 — Lead-in to the June cohort

23 June 2026 — Lead-in to the July cohort

Your facilitators

Two facilitators in every live session

Delivered by Thomas Byrnes and Marie-Josée Hamel. One facilitator leads the content while the other reads the room, supports participants who need more time, and stretches those who are ahead.

Thomas Byrnes

Thomas Byrnes

Lead facilitator

15+ years in humanitarian operations across 20+ countries, lead author of the GIZ DCI AI Hub Global Evidence Review, and the Rights-Based Risk Framework for AI in Social Assistance.

Marie-Josée Hamel

Marie-Josée Hamel

Operations Director, MarketImpact

Keeps the live cohort grounded in delivery, participant support, and operational realities so the practice transfers back into real teams.

Choose your format

Start, practise, build, or embed

Each format is practical on its own and gives you a clear next step when you are ready for more support.

Start

Online-only intro

A self-guided entry point for one safe starter workflow.

EUR 35

Lite Workflow Card

Register interest

Practise

Live Intro Lab

A facilitated introduction to responsible AI workflows.

EUR 75

Lite Workflow Card

Register interest

Build

Full six-session open cohort

Move from one-off prompting to responsible, repeatable workflows.

EUR 350 per person

EUR 280 for staff from national NGOs, local organisations, and aid workers between roles

Full Cohort Workflow Card

Apply for June cohort

Embed

Private team cohort

The full cohort adapted to your team workflows and risk boundaries.

Custom quote for teams

Full Cohort Workflow Card

Discuss private cohort

Full cohort detail

Six sessions over three weeks

The open cohort is for people who want to move from occasional AI use to responsible, repeatable workflows. Participants leave with a documented Workflow Card for one real recurring work process.

Short pre-training videos cover the basics before Session 1. Live sessions are reserved for practice, discussion, and guided failure, the moments where participants hit AI failure modes firsthand and learn to catch them. We do not waste instructor time on setup demos.

  • 6 live sessions x 90 minutes over 3 weeks
  • Recordings and between-session materials
  • CHEF Prompt Catalogue, reference materials, and verification prompts
  • Reusable Project or agent design
  • Full Cohort Workflow Card
  • Certificate of completion
  • Endline survey and post-course support path
  • Optional 45-minute follow-up call

What you leave with

Skills

  • CHEF prompting method for real tasks and source documents
  • Verification prompts and fresh-chat checking habits
  • Coach, Build, and Challenge agent-team design
  • Right-tool mapping across chat, Project, multi-agent, Cowork, and scheduled tasks
  • Risk review and human oversight plan

Artifacts

  • Reusable Projects or agents for recurring work
  • CHEF Prompt Catalogue and Workflow Card for one real recurring workflow
  • A one-month Claude Pro licence included in the course fee

Support

  • 30-day and 60-day check-ins to support sustained adoption

Course curriculum

The six-session pathway

Each session adds one working habit and ties the tool back to CHEF, verification, and human accountability.

Session 1

Get Past Generic AI

CHEF prompting with a real task and source document.

Output: First usable CHEF prompt, saved reflection entry, and first Prompt Catalogue row.

  • Introduce CHEF as the course method.
  • Show the difference between AI memory and supplied context.
  • Establish the habit of using real tasks, real documents, and real judgement.

Session 2

Breaking the First Answer

Adversarial audit and fresh-chat verification.

Output: Improved output plus verification prompts added to the Prompt Catalogue.

  • Treat first answers as useful drafts, not final truth.
  • Use same-chat adversarial audit and fresh-chat verification.
  • Name what each verification method catches and misses.
  • Participants run the verification technique on their own work and experience AI failure firsthand. That is how verification stops being theory and becomes habit.

Session 3

From Prompt to Project

Projects as saved CHEF prompts.

Output: First Project or agent for a recurring task, recorded as a reusable catalogue entry.

  • Move from one-off prompting to reusable Projects or agents.
  • Explain what persists: instructions, knowledge files, and chat history.
  • Introduce the Creator / Auditor split.

Session 4

Your Team of Agents

Coach, Build, Challenge; specific vs general agents; panel-of-experts demo.

Output: Agent-team map, Workflow Card architecture, and Coach / Build / Challenge catalogue entries.

  • Reframe Projects as agents using tool-agnostic language.
  • Introduce Coach, Build, and Challenge roles.
  • Participants choose their Workflow Card architecture.

Session 5

Working Across Files

Folder-based and multi-file work with permission risks.

Output: Low-stakes file-based test or documented readiness plan.

  • Compare chat, Projects, and desktop/file-based tools.
  • Discuss permissions, shared drives, accidental edits, deletion, and connected-app risk.
  • Make clear the core course does not depend on Cowork or desktop tools.

Session 6

Making It Stick

Choosing the right tool, risk review, and next steps.

Output: Final Prompt Catalogue, Workflow Card, Monday-morning commitment, and post-course support path.

  • Review CHEF and Coach / Build / Challenge.
  • Discuss personal, organisational, sector, and beneficiary risks.
  • Finalise the Workflow Card and operational checklist.

Workflow Card system

Lite and full cohort cards

The Workflow Card is the practical bridge between learning about AI and using it responsibly on real work.

Lite Workflow Card

Online-only and Live Intro Lab learners.

One real, low-risk task turned into a repeatable AI-assisted workflow: one task, one reusable prompt or agent, one verification step, and one practical next action.

  • One low-risk workflow candidate
  • One CHEF brief
  • One reusable prompt or primary agent
  • One verification habit

Full Cohort Workflow Card

Six-session open cohorts and private team cohorts.

A tested record of one real AI-assisted workflow, including the task, source material, agent design, verification method, risks, safeguards, and operational checklist.

  • Workflow architecture and agent-team design
  • Verification and human oversight plan
  • Risks, safeguards, and operational checklist
  • Personalised draft where enough detail is submitted by Session 4

Participants who submit enough workflow detail by Session 4 may receive a personalised draft Workflow Card. Others complete the standard card with facilitator support.

What participants say

Previous participants valued the practical workflow focus and realistic safe-use discussion.

Public feedback

"Key lessons from the training are definitely here to stay."

Lenka Blanarova, Head of Nutrition, Action Against Hunger UK

LinkedIn, April 2026

"Taking long reports, transcripts, or datasets and distilling them into clear summaries or bullet points saves hours of manual reading and note-taking."

Digital CFM Officer, NRC Sudan

"I really appreciated the training and found it practical."

Digital CFM Officer, NRC Sudan

"The AI training gave me the opportunity to learn new technology and improve the skills I use in my job responsibilities."

DCH Technical Assistant, NRC Sudan

"I thought two sessions a week was a lot for me, but after the first week, I knew those few sessions would help me save weeks of work in the long run. It was a great self-investment."

Moayad Zarnaji, MarketImpact

April 2026 cohort endline survey

Ready for the next facilitated cohort?

Apply for June cohort

Trusted by

Norwegian Refugee Council
Caritas Switzerland
International Rescue Committee
Estonian Refugee Council
Action Against Hunger UK

Not compliance training

This is not a compliance checkbox

Compliance training tells your staff what not to put in. This course teaches them how to catch what AI gets wrong. That is a different skill, a different format, and a different outcome.

Investment

Clear public pricing

Online-only and Live Intro Lab formats are coming soon. Register interest and we will follow up with the right next step.

Full six-session open cohort

EUR 350 per person

EUR 280 for staff from national NGOs, local organisations, and aid workers between roles

Six 90-minute live sessions over three weeks

Apply for June cohort

Private team cohort

Custom quote for teams

Six-session structure adapted to your organisation

Discuss private cohort

Eligibility for the reduced rate is confirmed at application stage based on the organisation or context you note in your application. There is no separate form.

Frequently asked questions

Do I need technical experience?

No. The course is designed for non-technical professionals. We focus on practical judgement, good prompts, verification, and safe workflows rather than coding.

Which format should I choose?

Choose the online intro for a low-cost start, the Live Intro Lab if you want facilitation, the full cohort if you want a supported workflow build, and the private cohort if your team needs adaptation to your tools and risk boundaries.

Who qualifies for the reduced rate?

Staff from national NGOs (national NGOs as defined by the country of registration, not the country of operation), staff from local organisations, and aid workers currently between roles. If you are unsure whether you qualify, apply for the standard rate and mention your situation in the application notes. We will confirm eligibility before invoicing.

What tools will we use?

The method works on any AI tool. We provide a one-month Claude Pro licence with the course fee (approximately EUR 20 value) because the Claude ecosystem (Projects, agents, file-based workflows) currently supports the widest range of what we teach in Sessions 3-6. Everything you learn transfers to ChatGPT, Copilot, Gemini, Mistral, or local open-weight models. We are not sponsored by any AI provider. If a better platform emerges for teaching these workflows, we will move to it.

What if my organisation does not allow Cowork or desktop tools?

Cowork and desktop/file-based tools are optional and only used where installation, file access, and policy allow it. The core course works with approved chat or Project-style environments.

Will I get a personalised Workflow Card?

Participants who provide enough workflow detail by Session 4 may receive a personalised draft. Everyone receives the standard Full Cohort Workflow Card and facilitator guidance.

What happens if I miss a session?

Recordings and materials are shared with participants. You can catch up between sessions and use the learner materials to keep your Workflow Card moving.