Seven hours of work in just over an hour: a language teacher compares Bruukki, ChatGPT and doing it by hand

A language teacher made a week of learning materials with Bruukki in just over an hour. Without AI, the same work took seven hours. Plus a ChatGPT comparison.

Jaakko Rannila, founder, Bruukki
Seven hours of work in just over an hour: a language teacher compares Bruukki, ChatGPT and doing it by hand

A language teacher with 12 years of experience used Bruukki to make one week of learning materials: a text, a vocabulary list of 81 words and phrases, and four tasks. It took one hour and five minutes. Making the same set without AI used to take seven hours.

The teacher recorded where the time went each time. The teacher also compared the result with how the same work had gone with ChatGPT. Below are the teacher’s observations.

From seven hours to just over one

The starting point was a 445-word article in a foreign language the teacher had used in class before. From it came a week’s set: a vocabulary list with Finnish and English translations, two vocabulary tasks, a reading comprehension task and a conversation task.

StepWithout AIWith Bruukki
Vocabulary (81 words and phrases)about 4 hoursabout 35 minutes
Tasksjust over 2 hours (3 tasks)20–30 minutes (4 tasks)
Whole set, including editing the textabout 7 hours1 hour 5 minutes

The work took about 85% less time. There was also one more task than in the original version. The comparison leaves out finding a suitable text. Finding this particular article once took the teacher about half an hour without AI. By the teacher’s own measurements, finding a suitable article before using AI sometimes took up to two hours. When working with Bruukki, the teacher already knew the text, which sped the work up somewhat.

Vocabulary: from four hours to 35 minutes

The vocabulary was the heaviest part. Without AI, the teacher picked the words and phrases, moved them into the list and translated them into Finnish and English. That took about four hours.

In Bruukki, the teacher picked the words by clicking them in the text, so nothing had to be copied by hand. Adding the words took about 20 minutes. Bruukki turned the words into their base forms as requested. The teacher corrected five of the 81.

The translations took into account the sentence each word appeared in. Only three translations out of 81 needed fine-tuning. The list also kept the original sentences, so students can see the context in which each word appears in the text. The whole list, translations included, was ready in about 35 minutes.

Tasks ready without corrections

The teacher asked for four tasks. Besides a phrase translation task, there was a gap-fill in which the phrases are used in sentences Bruukki wrote. A reading comprehension task and a conversation task completed the set. They took 20–30 minutes, where three tasks had taken just over two hours before.

“The tasks needed no corrections at all.”

Bruukki also added things the teacher had not asked for: task instructions, solutions for the teacher and tips for students in the conversation task. There was no need to repeat the proficiency level in every request. About the conversation task, the teacher wrote:

“The conversation task was even better than the one I had made myself.”

Compared with ChatGPT and Claude Cowork

The teacher had made tasks with ChatGPT before. The experience was different:

“When I used to make tasks with ChatGPT, I could use half of what the chat produced at most, if even that. ChatGPT produced a lot of irrelevant content.”

The same showed in translations. In the teacher’s experience, ChatGPT makes many inaccurate translations and does not always take the sentence into account. In a vocabulary list, the sentence is exactly what decides which translation is right.

In another case, the teacher made an A1-level foreign-language lesson set twice: with Bruukki and with the general-purpose AI agent Claude Cowork. With Bruukki the set was ready in eight and a half minutes, with Cowork in 25 minutes. About 80% of Bruukki’s result was usable as it was, compared with about 70% of Cowork’s. Bruukki also included a listening task, which Cowork could not produce as a ready task.

Other teachers have compared it too

Other teachers have also compared Bruukki with other AI tools.

“With Bruukki you can make better learning materials than with Copilot Premium.”

Mathematics teacher, vocational upper secondary

“Bruukki’s AI can do things that two other AI tools I’ve tried could not.”

Mother tongue teacher, comprehensive school, 20+ years of experience

Two vocational upper secondary teachers compared Bruukki with their institution’s own Copilot. In their experience, Copilot forgets midway through what you were working on. In Bruukki, the work continues where you left off. Bruukki also uses the background the teacher gives, such as age group and field, as part of the task instead of tacking it on separately.

Why the difference

With a general-purpose AI tool, the teacher states the proficiency level, the task type and the text in every request. Then they cut whatever does not fit their teaching.

Bruukki is a pedagogical AI. It already knows the structures of teaching, such as proficiency levels and task types, and builds the tasks from the text in the same material. The vocabulary, the tasks and the listening all go into one material, and the teacher edits or rejects every suggestion. Bruukki also knows schools, curricula and their course units, for example the Finnish national and local curricula. That makes it easier to create material for a specific course unit, because Bruukki takes the unit’s objectives, content and assessment criteria into account.

Bruukki also learns what the teacher wants. From recent chats it infers things like the teacher’s teaching style, content tone and preferred length. The more often you use Bruukki, the better its suggestions fit. Teachers can see these inferred settings in their profile and edit or reset them.

With an institutional licence, the school or college can also save its own priorities and guidelines in Bruukki as skills. Bruukki then makes plans and learning materials within the framework the institution has set. A guideline can cover, for example, competence-based tasks, assessment criteria or taking student workload into account.

Time is saved, and the quality of the learning material stays in the teacher’s hands.

Try it with your own text

Take a text you already use in your teaching, pick the words and ask for the tasks. Try Bruukki for free.

Read also: differentiation in language teaching and support for the whole teaching week.

Bruukki's AI was used to help write this blog post.

Try it with your own material

The AI suggests, you decide.