If I could choose, I would wish for every person in conflict to have a caring, experienced mediator by their side. Someone who listens, who asks the right questions, who holds space. Someone made of flesh and blood.
And yet I work full-time on building an AI-supported mediation platform. This text is an attempt to describe that tension without pretending to resolve it.
A Question of Access
As a political scientist, I learned to see conflict not as a disturbance but as a normal part of living together in society. Where people with different interests, needs, and strategies meet, conflicts arise. The decisive question is therefore not whether or why they exist, but how we deal with them. Along the way, I learned to hold two perspectives at once: to see the numbers behind the individual stories — and the people within the numbers.
We all know someone who was on sick leave because of a conflict at work, or who in the end even changed jobs. As early as 2009, the KPMG conflict cost study put the cost of conflict in companies at 15 to 20 percent of annual personnel costs. Many companies and organizations do notice conflicts — but have no effective ways to address them early and constructively.
At the same time, civil court proceedings have been declining significantly for years. That does not automatically mean that people are turning to other ways of resolving conflict more often. The more pressing question is what happens to the conflicts that never find their way into judicial or professional hands. The high costs of workplace conflict and Germany’s above-average sickness absence rates do not suggest that conflicts are disappearing — rather, that they often remain unaddressed.
From my own practice, I know this gap very concretely. I regularly recommend professional conflict support and see, at the same time, how rarely it is actually taken up. Colleagues report the same. The capacity is there. The problem is often not the supply but the access — everything that keeps people from even considering professional support: uncertainty, shame, lack of information, or the sense that the conflict isn’t “bad enough” yet.
In political science terms, it is therefore not enough to ask what options for addressing conflict exist. You also have to ask who has access to them — and who doesn’t. Mediation makes this especially clear: many people never opt for professional support, regardless of whether they know what mediation is. The threshold remains high.
So one question won’t let me go: what happens to all the conflicts that never find their way to constructive resolution? For me, one answer is to use every possibility that can lead people into constructive support earlier and more reliably — through public discourse, through counselling services, and, yes, through technology.
Technology and Mindset
My relationship with technology is not uncritical. As a researcher, I observe how technological developments change social structures. Often not for the better. Historically, though, the problem was rarely progress itself, but rather the mindset with which we used it. And, again, the question of who had access: to new tools, to knowledge, to the possibilities that grew out of them.
This shapes my private decisions too. My children are nine years old; they have no smartphone and no social media. But they have their own computers and are learning to use the internet and media consciously. Not avoidance, but shaping.
I carry this mindset into my work with AI. Language models can change the way people talk to one another, the way they deal with conflict. Technology can isolate or connect. It can help people hide from conflict, or make working through it easier. Which direction it takes depends on how it is designed. I want AI to enable people to cooperate. That is why I use it to bring people closer together.
What AI Cannot Do
The objections to AI in mediation are legitimate. They run through this project like a red thread.
AI is not capable of real empathy. That’s true. A language model simulates understanding, but it does not understand. It has no experience of its own — of hurt, fear, or hope. This limit is fundamental, and it determines what AI can be in conflict resolution: a tool, not a relationship.
My co-founder Aline described in a blog post how, in our first tests, the AI immediately proposed solutions and offered compromises. It was maximally helpful — and in doing so violated the basic principles of mediation as we largely understand them in Germany. Mediators do not prescribe solutions. They do not judge. They provide structure. That is hard for us humans. For an AI whose core instinct is to help, it is even harder.
We had to work intensively to guide the AI toward restraint. To teach it that conflicts rarely unfold in a straight line, because insight sometimes simply takes time. And this work is far from finished. It also makes this project a different kind of technical challenge from most AI applications: we are not optimizing for efficiency, but for the quality of the process.
What AI Might Be Able to Do
Despite these limits, I see a possibility for AI as a door-opener.
Many people affected by conflict do not initially seek professional support. The threshold is too high, the situation doesn’t seem bad enough yet, the right moment passes. By the time support is sought, much trust has often already been destroyed.
An AI could come in at exactly this point: as a low-threshold entry that helps people sort out their situation. As structured reflection that prepares them for a difficult conversation. As a first step that paves the way to human support.
For that to work, we have to design the AI so that it raises no false expectations and hands over to humans in time when the situation requires it.
How AI Can Strengthen Mediators
Mediators are so much more than mere “providers of structure” or “conflict guides.” They are the ones who — at the moment when we want to see only the worst in the other person, and often enough that other person is someone we were once very close to — bring us back to our positive humanity. They are often the first and only ones to ask: “What do you need?”
My standard, therefore, is that AI support should give us mediators time for what really counts: the work with people in complex situations. Documentation, preparation, and administrative tasks — those I am happy to have taken off my plate. At the same time, AI can also be helpful where practice and confidence are concerned: as a space for reflection, as a training partner, as a way to play through difficult conversational situations in advance and to try out one’s own way of leading a conversation more deliberately. Creating a protected space and enabling understanding, however, remain human core competencies for me.
And if AI helps more people find their way to mediation in the first place, then to me that means: more constructive conflict culture, more human encounter.
Confidentiality as the Foundation
Mediation lives on confidentiality. Without the trust that what is said stays protected, the process does not work. This principle applies to AI-supported tools to a special degree.
That is why confidentiality has to be part of the design from the very beginning — technically, organizationally, and conceptually. For us, that means developing AI in this field with a clear focus on confidentiality, security, and data protection. That is demanding. But it is the precondition for AI being used responsibly in mediation at all.
No Conclusion — an Invitation
I want to use every possibility to give more people access to constructive conflict resolution. Earlier, at a lower threshold, and on a path that leads toward human support — not away from it. What AI can really contribute here will only become clear in practice. I am ready to learn from that practice and to keep helping shape it. If these questions are on your mind — as a mediator, as a leader, as a person in a conflict — I would be glad to hear from you. Get in touch anytime.
Julia Stadler is co-founder and CEO of Tuplita. She holds a PhD in political science and is a teacher and certified mediator (member of the Bundesverband Mediation, the German Federal Association for Mediation).
At tuplita.ai, we build with confidentiality, security, and GDPR compliance as foundational requirements.
