Efficient cooperation integrating AI
2024 · Bergen · 2 min read
Team Agnes Villmones Haug, Anya Røng Skatvedt, Arian Velázquez, Frida Ege Olsen and Ole Sander Brudvik
Read more about the evaluation of BriefBot Test the prototype here!
- Problem
- Information overload, poor email management and ineffective communication inside one of Norway’s largest media organisations.
- Approach
- Semi-structured interviews and thematic analysis, brainstorming, personas, storyboards and paper wireframes; usability tested in Maze.
- Outcome
- An AI assistant that automates calendar handling and morning reports and works across the tools the newsroom already uses.
BriefBot is an AI assistant for a newsroom drowning in email — built with one of Norway’s largest media organisations, and grounded in the field of Computer Supported Cooperative Work.
Hundreds of emails a day
Semi-structured interviews, analysed thematically: every point on a digital post-it in FigJam, sorted into communication, collaboration and AI, then into seven themes. The picture was consistent — information overload, poor communication, time wasted, messages missed. One department head described hundreds of emails a day and deleting them to cope, with the stress and missed deadlines that follow. The leader of the EAI team wanted AI on it.
From post-its to paper
- Brainstorming, then “How might we…” — problems turned into openings.
- Four personas, for the range of people who would use it.
- One storyboard: an imaginary day with it.
- A great many paper wireframes, to decide what mattered.
BriefBot
BriefBot is an AI assistant that takes the mundane off the newsroom’s plate and works inside the tools already in use.
- It learns the job. A thorough registration teaches it the role; it keeps learning from how you work, and the home screen is yours to arrange.
- A calendar that keeps itself. Meetings and deadlines land in it on their own, so nothing is lost in the flood.
- Morning reports. A brief every morning: today’s tasks and deadlines, summarised.
- Slack and mail, tamed. Long threads summarised, emails translated and sorted by importance. Pick the apps to include, and it works across them.
- Accessible. Font sizes, voice commands, a text-to-speech reader, colourblind and dyslexic modes, light and dark — built against WCAG, with full compliance the aim.
Tested in Maze
- Usability tested in Maze.
- Time per task, feedback after each — quantitative and qualitative in one pass.
- The write-up is linked at the top.