The Sand Wolf

About the project

The Gothenburg Opera had the idea of an AI chatbot based on the character the Sand Wolf from the children's books and the opera of the same name. Our friends at &friends taught an AI to talk to the children in the Sand Wolf's way, and we were responsible for the development.

services

Communication

client

The Gothenburg Opera

Two mobile phones showing a chat with Sandvargen on opera.se

Dackadida!

The AI chatbot answers children's questions about big and small things in the clever and ingenious way that only the Sand Wolf can. Together with &friends, we created a personalized user interface.

About the technology

The public sector cannot use AI with servers in the US. Therefore, we created a widget for the chatbot using React/NextJS, which in turn communicated with a proprietary API layer. We hosted the entire solution within the EU via the company Scalingo, which was simple, cost-effective, and fully in line with the City of Gothenburg's data and security policies. Our API, in turn, communicated with a Chatbase bot, which our colleagues at &friends had trained to behave exactly like the Sand Wolf.

FAQ

Can an AI assistant you build make up answers?

No. The assistant answers only from your own material and links to the source. If it cannot find an answer, it says so.

Technically, it searches your documents for the right passages and writes the answer from them. It is not allowed to answer freely from the language model's own memory.

Because every answer shows where it came from, users can check it. If something is wrong, it can be traced to a document and corrected there.

Can an AI assistant you build make up answers?

No. The assistant answers only from your own material and links to the source. If it cannot find an answer, it says so.

Technically, it searches your documents for the right passages and writes the answer from them. It is not allowed to answer freely from the language model's own memory.

Because every answer shows where it came from, users can check it. If something is wrong, it can be traced to a document and corrected there.

Can an AI assistant you build make up answers?

No. The assistant answers only from your own material and links to the source. If it cannot find an answer, it says so.

Technically, it searches your documents for the right passages and writes the answer from them. It is not allowed to answer freely from the language model's own memory.

Because every answer shows where it came from, users can check it. If something is wrong, it can be traced to a document and corrected there.

What happens to our data when you build an AI service for us?

It stays yours. You decide where content is stored and who can read what, and the solution complies with GDPR from the start.

Content stays in your own environment or with a vendor you have approved, within the EU. We do not train any AI models on your material.

The assistant follows the permissions you already have. Anyone who is not allowed to read a document will not get answers based on it either.

What happens to our data when you build an AI service for us?

It stays yours. You decide where content is stored and who can read what, and the solution complies with GDPR from the start.

Content stays in your own environment or with a vendor you have approved, within the EU. We do not train any AI models on your material.

The assistant follows the permissions you already have. Anyone who is not allowed to read a document will not get answers based on it either.

What happens to our data when you build an AI service for us?

It stays yours. You decide where content is stored and who can read what, and the solution complies with GDPR from the start.

Content stays in your own environment or with a vendor you have approved, within the EU. We do not train any AI models on your material.

The assistant follows the permissions you already have. Anyone who is not allowed to read a document will not get answers based on it either.

Do AI assistants work in the public sector and regulated environments?

Yes, those are the environments we build for. Security, data handling and the requirements of the EU AI Act are handled from the start.

Before we build anything, we go through where data may be stored, which vendors you have approved and how access should be controlled. Hosting within the EU is the starting point.

The AI Act requires the system to be classified correctly, documented and owned by someone. We do that part together with you, since much of it is about your own routines.

Do AI assistants work in the public sector and regulated environments?

Yes, those are the environments we build for. Security, data handling and the requirements of the EU AI Act are handled from the start.

Before we build anything, we go through where data may be stored, which vendors you have approved and how access should be controlled. Hosting within the EU is the starting point.

The AI Act requires the system to be classified correctly, documented and owned by someone. We do that part together with you, since much of it is about your own routines.

Do AI assistants work in the public sector and regulated environments?

Yes, those are the environments we build for. Security, data handling and the requirements of the EU AI Act are handled from the start.

Before we build anything, we go through where data may be stored, which vendors you have approved and how access should be controlled. Hosting within the EU is the starting point.

The AI Act requires the system to be classified correctly, documented and owned by someone. We do that part together with you, since much of it is about your own routines.

More cases

All cases

PubQ's homepage on a tablet on a wooden table, next to a pen

Checkout on the homepage

PubQ

Curious minds

University of Borås

Crossing the tram tracks on a tree-lined street in Gothenburg

What gives you energy?

Göteborg Energi

Checkout on the homepage

PubQ

Curious minds

University of Borås