AI that serves support agents: Zammad’s Product Owner on balancing innovation and user autonomy

From Consultant to Product Owner – what new insights have you gained into Zammad through this shift in roles?

My journey with Zammad started as a user – back then, I knew the product purely from a customer perspective. Later, I became a consultant, and today, I’m the Product Owner. With each role, my understanding of just how much effort goes into individual features has deepened.

As a user, I saw Zammad as hands down the best open source helpdesk solution. As a consultant, I was impressed by the wide range of use cases Zammad could support. My job was to get the most out of the system for each client – and I often thought, “This little feature? That should be a quick win.”

“As a consultant, I developed solutions for individual customers. Today, I evaluate whether and how these needs can be meaningfully integrated into the product – in a way that is sustainable and benefits many users.”

Now, I realize there’s a lot more to a feature than just building it. It has to be implemented cleanly, tested, documented, and integrated into the system’s architecture. Most importantly, it needs to make sense for many users – not just one specific use case.

The questions I ask myself now as a Product Owner are different: How can we turn real user needs into sustainable, scalable features? How do we design solutions that truly add value? That shift in thinking has been one of the biggest changes in perspective for me.

What challenges have you faced in your new role so far?

The biggest challenge right now is balancing technical modernization with ongoing feature development. We’re investing heavily in updating our tech stack – but at the same time, the market expects innovation, especially around AI.

“We are renewing the foundations while already working on the top floor. This balancing act is what my role requires right now.”

In an ideal world, we’d focus entirely on the desktop rewrite. That would be the most efficient path. But the market doesn’t wait. So we have to move forward on both fronts at once.

To make that work, we’re extremely deliberate in how we prioritize. New features are planned to run on both stacks. We put a strong focus on reusability and clean interfaces – which takes coordination, but ultimately pays off.

What can Zammad users look forward to in the coming months?

We’re working on AI-powered features – but we’re approaching the topic with care. Our focus is on using this technology responsibly and ethically.

From a technical standpoint, many things are easy to implement. But not everything that’s possible delivers reliable results in actual practice. Take AI-generated text, for example: hallucinations may be less common now, but they’re still very much a challenge.

I actually ran a little experiment myself: over several months, I deliberately contacted the support teams of various companies – even with trivial questions. The experience was sobering. Far too often, I received AI-generated responses that completely missed the point. It often took multiple emails before a human stepped in and had to untangle the mess. Frustrating – and eye-opening.

“We’re building AI features – but not at any cost. Responsibility and quality come first.”

At Zammad, we're taking a different approach. Our AI is designed to support agents - not replace them. For example, it can suggest responses, but the final decision about what goes out to the customer is still made by humans. That's how we increase efficiency without compromising the quality of customer communication.

We’re currently experimenting with several AI-driven features. Which ones will make it into the next release is still to be determined:

• Automatic ticket categorization: Incoming requests are pre-sorted and routed to the right teams – without relying on complex, manually maintained rules.
• Concise ticket summaries: Long conversations are reduced to the essential information.
• Seamless translation support: Multilingual support teams can work more efficiently without needing external tools.
• Context-based response suggestions: Our AI assistant analyzes ticket content and proposes replies – factoring in knowledge base content and internal documentation.

What makes Zammad’s approach to AI different from closed systems?

Zammad has always been an open product – not just because the source code is publicly available, but also thanks to our powerful APIs. And we’re carrying that same commitment to openness into how we approach AI.

At the heart of it is choice. Zammad users will be able to decide for themselves which AI model they want to use. After all, AI features are made up of two core components: the user-facing interface – such as ticket summaries or response suggestions – and the model working in the background that powers those features.

“We don’t chase trends. We build with substance and responsibility. That’s what makes our open source model sustainable.”

We’re deliberately building for flexibility. Whether it’s OpenAI, Claude, Mistral, or a self-hosted model like Ollama – we want to make it as easy as possible to integrate a wide range of solutions. Especially for on-premise users, AI introduces new technical challenges. Unlike traditional systems, AI often requires hardware like GPUs or NPUs – something many standard servers aren’t equipped for. Some teams prefer to run their own infrastructure, others opt for external providers.

We’ll also be offering an option for those who want a fully managed solution: SaaS customers will then be able to add Zammad’s hosted AI services to their setup. In that case, Zammad GmbH will take care of both the helpdesk system and the underlying AI model – a convenient all-in-one solution for anyone who wants to keep things simple.

And how does Zammad position itself compared to other ticketing systems with AI features?

The key difference is freedom of choice. While many helpdesk providers offer AI features as a fixed, black-box integration, Zammad puts users in control. You choose which model runs in the background – whether it’s self-hosted, connected via a third-party provider, or managed through Zammad itself.

Because when it comes to AI, the questions aren’t just technical. They’re also about privacy, ethics, and trust. Other platforms may claim to take those issues seriously – but how exactly they handle them is often unclear, or buried deep in the fine print.

“Openness isn’t just a detail – it’s a core principle. At Zammad, users decide which AI model runs in their environment.”

That approach matters – especially in the world of AI. Many models were trained on content with unclear origins, sometimes without the consent of the original creators. With Zammad, users stay in control: of the technology they use, of how their data is handled, and of the ethical standards they choose to uphold.

That’s why our position is clear: Zammad doesn’t just offer AI features – we offer the freedom to integrate AI on your own terms, in a way that fits your organization, your privacy needs, and your principles.

What excites you personally about Zammad’s future?

I’m convinced that Zammad is already the best open source ticketing system on the market. And yet, it still feels like we’re just getting started.

With the new tech stack and upcoming AI capabilities, we’re on the verge of a real leap forward. What excites me most is that we’re not chasing trends – we’re building solutions with substance: thoughtful, meaningful, and with real value for our users.

“Zammad is already strong – but we’re only at the beginning. What’s coming next will take the product to a whole new level.”

When I look ahead, I see Zammad becoming the leading open source alternative to the big, closed systems. Our mission has always been – and continues to be – to offer a support platform that is responsible, technically advanced, and genuinely user-friendly. One where innovation and openness aren’t in conflict – but go hand in hand.

That mindset runs through the entire team – and it’s exactly what makes working on Zammad so inspiring.