AI Agents
- Within our AI solution, we are exploring the potential of AI agents.
- AI agents are smart systems with the ability to act independently.
- This means they can perform actions, collaborate with other agents, and use different tools to achieve objectives, while also being able to plan their work.
- Instead of detailed, step-by-step instructions, the user only needs to define the goal and context and approve important decisions.
Human-Centered DNS Agent
We aim to explore how to address the challenges with Registration Data through Human-Centered AI principles. Human-Centered AI (HCAI) is about designing AI systems that enhance human abilities rather than replace them. For example:
- AI suggests, Human Decides.
- Transparency – AI must clearly explain how conclusions are reached.
- Users must be able to adjust and supervise the AI’s behaviour.
- AI tools improve over time as they learn from data — but good design lets users guide and speed up that learning. We call this feedbackloops.
- We also want to consider GDPR’s principle of Privacy by Design, which means building in data protection from the start (—by minimizing data use, securing it, and giving users control.)
Conversations complements navigation
Traditional data management systems rely on navigation through menus and forms – this model feels a bit outdated today. So a third design focus was to explore a more direct way to work with data.
A conversational user interface would let users ask questions and get answers — visualizing tables with only the relevant information designed to be interactive. We also tested a hybrid design that allows users to switch between conversation and navigtion.
Testing and Iteration
We tested the prototype and got a lot of valuable feedback and iterated the design.
Some things worked very well — like…
- The conversational interface, which felt fast and easy to use.
- The hybrid version was also appreciated for giving users more control, depending on the task.
- Issue alerts on inaccurate registration data in the domain name portfolio were seen as highly valuable. These issues are often only found when they start causing trouble.
We also got constructive feedback.
- One concern was traceability — users wanted to understand where issue alerts came from. So, we made sure the source could be easily found behind each alert.
- Another request was the ability to track and undo changes — So we added a detailed log and an undo function.