Skip to content
A five-step workflow for one programme description: last year's text, an AI draft from the master prompt, staff edit the draft, translation into Finnish, Swedish and English, then a check by marketing

University of Helsinki · 2025

One standard, three languages

AI prompts that draft and translate degree programme descriptions for Studyinfo, in Finnish, Swedish and English.

Client
University of Helsinki
Year
2025
What we did
Workflow, Localisation

The University of Helsinki's marketing team asked us to help faculties write better degree programme descriptions for Studyinfo, the national site where students compare and apply for programmes. We wrote and tested the AI prompts. The marketing team and the university's AgentFormer project supplied the platform.

About 100 programmes, three languages

Each degree programme writes its own description, and the marketing team brings them together. With about 100 programmes, the texts varied in quality, structure and tone. Each one also has to work in Finnish, Swedish and English, and the university's official terms have to stay the same in all three.

The solution had to suit academics, who rightly check every word, and it had to keep the people who know the programme in charge of the facts.

Who did what

MXM
Wrote and tested all the prompts: drafting, tone and translation.
Marketing team
Model texts, tone of voice, the list of official terms, and testing.
AgentFormer
The university's own project. It supplied the platform built to run the prompts.

The prompts

  • A master prompt that turns a programme's existing text into a draft in the Studyinfo structure, in British English. Bachelor's texts are more encouraging, master's texts more formal, and neither may sound like advertising.
  • Translation prompts for Finnish, Swedish and English, written as an academic translator would work. They check their own result before answering and use the university's official terms as a fixed list.
  • Rules from testing. The career section never promises a job, only the skills a student gains and possible career paths. Facts such as the language of teaching stay as the programme staff wrote them; only the language is corrected.

The workflow we designed

In our design, the marketing team would send each programme a package: last year's text, an AI draft and the master prompt. Programme staff would edit the draft, the prompts would translate it, staff would approve the result, and marketing would check and format it for every channel. The final text would become the starting point for next year.

We tested the prompts on real programme texts, starting with the bachelor's programme in Geography, and compared models from OpenAI and Google on the same texts before choosing.

Do many people in your organisation write the same kind of text? Talk to us.

Want this for your team?