Getting started with AI
for nonprofits and social organisations.

There is plenty of research and teaching about artificial intelligence in social work. Less often does someone turn up and build something. That is exactly what I do — with seven years of my own practice in employment integration behind me.

I know both sides of the table

Most people selling you AI have never sat in a counselling office. Most people who understand your work do not build software. I do both — and that is the only reason I am writing this page at all.

Seven years in employment integration

Work-related education, learning coaching, social counselling. I know what a day inside an institution looks like, where the time runs out and why well-meant tools end up unused.

Building a blended learning platform

From concept to daily operation — for job seekers, not for a pilot project. Details further down.

And I build things myself

Apps, websites, internal tools. If it turns out you need something that does not exist yet, I can develop it instead of referring you elsewhere.

Three stations, honestly placed

No case studies with invented percentages. Three stations I actually went through — with what they were and what they were not.

Blended learning platform for job seekers

Built and run within employment integration. Four areas that previously ran on lists, conversations and experience:

  • Course selection — which course suits whom, by criteria you can follow rather than by gut feeling.
  • Learning type assessment — people learn differently; the programme had to be able to reflect that.
  • Course progress — visible to the participants themselves, not only to their support workers.
  • Course evaluation — feedback that actually feeds into the next round.

The hard part was never the technology. It was deciding which judgements a tool may take over and which have to stay with a person.

SOREVA — one hub instead of five separate tools

There are plenty of solutions for individual areas. Each one is missing something different — and anyone running several ends up maintaining systems that know nothing about each other. That was the trigger.

SOREVA brings intake, participant administration and training together in one place — for institutions, public bodies, small businesses and independent consultants, with a focus on nonprofits and employment integration. Currently in development.

Asylum coordination, from 2011 — advising and building, together from the start

Social counselling, supporting asylum seekers and refugees, liaising with authorities. Even then I was building websites and digitalising workflows alongside it — with whatever was technically possible in 2011.

None of that has anything to do with AI, and I do not sell it that way. It is here because it shows that advising and building are not two phases of a career for me, but the same approach since 2011: look closely, work out what is missing — and then build it.

Four ways in

Depending on where you stand. It usually begins with the first one and turns out along the way to be the third.

Taking stock

Half a day: where you lose time today, what could sensibly be automated, and what had better not be. The result is an assessment, not a list of tools.

Workshop for the team

Hands-on, using your own tasks rather than textbook examples. For people without a technical background — which is the normal case, not the exception.

Support during operation

A workshop fizzles out if nobody is there afterwards. So I offer support over several months — with fixed appointments and the option to ask in between.

Building tools

If it turns out the right thing does not exist: I develop it. From a small internal tool to a full app — with the same person who did the consulting.

Workshops1:1 BegleitungProzessanalyseDatenschutzBlended LearningClaudeChatGPTGemini

What AI does not do in your field

This belongs at the front, not in the small print. People in social work who are sceptical usually have good reasons.

It does not make decisions about people

No model decides who gets a measure and who does not. It can summarise documents and make suggestions — the responsibility stays with the person who knows the case. Anything else would be untenable, professionally and legally.

It does not replace relationship work

What happens in a conversation cannot be automated. AI can free up time spent on administration — more time for conversations is the point, not less.

It is not automatically compliant with data protection

Client data does not belong in a chat window unchecked. Which tools qualify under which conditions is part of every consultation — and sometimes the answer is: none, for this purpose.

It does not repair broken processes

An unclear process only becomes unclear faster through automation. If it turns out your real problem is not a technical one, I will say so — even when that means no engagement comes of it.

From the first call to the first result

First call

30 to 45 minutes, free of charge, by video or in person in Basel. We see whether it fits at all.

Taking stock

Half a day to a full day with you. I listen, look at how things run and ask a lot of questions.

Assessment

In writing, in plain language, with a recommendation — including the recommendation to do nothing for now.

Putting it into practice

Workshop, ongoing support, a tool that gets built — or a combination. What comes of it is your decision.

If that sounds like your situation

The first call is free and without obligation. I sell nothing in it — we look at whether I am the right partner. If I am not, I say so.